drivers/net/ethernet/intel/i40e/i40e_main.c

   1 /*******************************************************************************
   2  *
   3  * Intel Ethernet Controller XL710 Family Linux Driver
   4  * Copyright(c) 2013 - 2015 Intel Corporation.
   5  *
   6  * This program is free software; you can redistribute it and/or modify it
   7  * under the terms and conditions of the GNU General Public License,
   8  * version 2, as published by the Free Software Foundation.
   9  *
  10  * This program is distributed in the hope it will be useful, but WITHOUT
  11  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
  12  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
  13  * more details.
  14  *
  15  * You should have received a copy of the GNU General Public License along
  16  * with this program.  If not, see <http://www.gnu.org/licenses/>.
  17  *
  18  * The full GNU General Public License is included in this distribution in
  19  * the file called "COPYING".
  20  *
  21  * Contact Information:
  22  * e1000-devel Mailing List <e1000-devel@lists.sourceforge.net>
  23  * Intel Corporation, 5200 N.E. Elam Young Parkway, Hillsboro, OR 97124-6497
  24  *
  25  ******************************************************************************/
  26
  27 /* Local includes */
  28 #include "i40e.h"
  29 #include "i40e_diag.h"
  30 #if IS_ENABLED(CONFIG_VXLAN)
  31 #include <net/vxlan.h>
  32 #endif
  33 #if IS_ENABLED(CONFIG_GENEVE)
  34 #include <net/geneve.h>
  35 #endif
  36
  37 const char i40e_driver_name[] = "i40e";
  38 static const char i40e_driver_string[] =
  39                         "Intel(R) Ethernet Connection XL710 Network Driver";
  40
  41 #define DRV_KERN "-k"
  42
  43 #define DRV_VERSION_MAJOR 1
  44 #define DRV_VERSION_MINOR 4
  45 #define DRV_VERSION_BUILD 8
  46 #define DRV_VERSION __stringify(DRV_VERSION_MAJOR) "." \
  47              __stringify(DRV_VERSION_MINOR) "." \
  48              __stringify(DRV_VERSION_BUILD)    DRV_KERN
  49 const char i40e_driver_version_str[] = DRV_VERSION;
  50 static const char i40e_copyright[] = "Copyright (c) 2013 - 2014 Intel Corporation.";
  51
  52 /* a bit of forward declarations */
  53 static void i40e_vsi_reinit_locked(struct i40e_vsi *vsi);
  54 static void i40e_handle_reset_warning(struct i40e_pf *pf);
  55 static int i40e_add_vsi(struct i40e_vsi *vsi);
  56 static int i40e_add_veb(struct i40e_veb *veb, struct i40e_vsi *vsi);
  57 static int i40e_setup_pf_switch(struct i40e_pf *pf, bool reinit);
  58 static int i40e_setup_misc_vector(struct i40e_pf *pf);
  59 static void i40e_determine_queue_usage(struct i40e_pf *pf);
  60 static int i40e_setup_pf_filter_control(struct i40e_pf *pf);
  61 static void i40e_fill_rss_lut(struct i40e_pf *pf, u8 *lut,
  62                               u16 rss_table_size, u16 rss_size);
  63 static void i40e_fdir_sb_setup(struct i40e_pf *pf);
  64 static int i40e_veb_get_bw_info(struct i40e_veb *veb);
  65
  66 /* i40e_pci_tbl - PCI Device ID Table
  67  *
  68  * Last entry must be all 0s
  69  *
  70  * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
  71  *   Class, Class Mask, private data (not used) }
  72  */
  73 static const struct pci_device_id i40e_pci_tbl[] = {
  74         {PCI_VDEVICE(INTEL, I40E_DEV_ID_SFP_XL710), 0},
  75         {PCI_VDEVICE(INTEL, I40E_DEV_ID_QEMU), 0},
  76         {PCI_VDEVICE(INTEL, I40E_DEV_ID_KX_A), 0},
  77         {PCI_VDEVICE(INTEL, I40E_DEV_ID_KX_B), 0},
  78         {PCI_VDEVICE(INTEL, I40E_DEV_ID_KX_C), 0},
  79         {PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_A), 0},
  80         {PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_B), 0},
  81         {PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_C), 0},
  82         {PCI_VDEVICE(INTEL, I40E_DEV_ID_10G_BASE_T), 0},
  83         {PCI_VDEVICE(INTEL, I40E_DEV_ID_10G_BASE_T4), 0},
  84         {PCI_VDEVICE(INTEL, I40E_DEV_ID_20G_KR2), 0},
  85         {PCI_VDEVICE(INTEL, I40E_DEV_ID_SFP_X722), 0},
  86         {PCI_VDEVICE(INTEL, I40E_DEV_ID_1G_BASE_T_X722), 0},
  87         {PCI_VDEVICE(INTEL, I40E_DEV_ID_10G_BASE_T_X722), 0},
  88         {PCI_VDEVICE(INTEL, I40E_DEV_ID_20G_KR2), 0},
  89         {PCI_VDEVICE(INTEL, I40E_DEV_ID_20G_KR2_A), 0},
  90         /* required last entry */
  91         {0, }
  92 };
  93 MODULE_DEVICE_TABLE(pci, i40e_pci_tbl);
  94
  95 #define I40E_MAX_VF_COUNT 128
  96 static int debug = -1;
  97 module_param(debug, int, 0);
  98 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
  99
 100 MODULE_AUTHOR("Intel Corporation, <e1000-devel@lists.sourceforge.net>");
 101 MODULE_DESCRIPTION("Intel(R) Ethernet Connection XL710 Network Driver");
 102 MODULE_LICENSE("GPL");
 103 MODULE_VERSION(DRV_VERSION);
 104
 105 /**
 106  * i40e_allocate_dma_mem_d - OS specific memory alloc for shared code
 107  * @hw:   pointer to the HW structure
 108  * @mem:  ptr to mem struct to fill out
 109  * @size: size of memory requested
 110  * @alignment: what to align the allocation to
 111  **/
 112 int i40e_allocate_dma_mem_d(struct i40e_hw *hw, struct i40e_dma_mem *mem,
 113                             u64 size, u32 alignment)
 114 {
 115         struct i40e_pf *pf = (struct i40e_pf *)hw->back;
 116
 117         mem->size = ALIGN(size, alignment);
 118         mem->va = dma_zalloc_coherent(&pf->pdev->dev, mem->size,
 119                                       &mem->pa, GFP_KERNEL);
 120         if (!mem->va)
 121                 return -ENOMEM;
 122
 123         return 0;
 124 }
 125
 126 /**
 127  * i40e_free_dma_mem_d - OS specific memory free for shared code
 128  * @hw:   pointer to the HW structure
 129  * @mem:  ptr to mem struct to free
 130  **/
 131 int i40e_free_dma_mem_d(struct i40e_hw *hw, struct i40e_dma_mem *mem)
 132 {
 133         struct i40e_pf *pf = (struct i40e_pf *)hw->back;
 134
 135         dma_free_coherent(&pf->pdev->dev, mem->size, mem->va, mem->pa);
 136         mem->va = NULL;
 137         mem->pa = 0;
 138         mem->size = 0;
 139
 140         return 0;
 141 }
 142
 143 /**
 144  * i40e_allocate_virt_mem_d - OS specific memory alloc for shared code
 145  * @hw:   pointer to the HW structure
 146  * @mem:  ptr to mem struct to fill out
 147  * @size: size of memory requested
 148  **/
 149 int i40e_allocate_virt_mem_d(struct i40e_hw *hw, struct i40e_virt_mem *mem,
 150                              u32 size)
 151 {
 152         mem->size = size;
 153         mem->va = kzalloc(size, GFP_KERNEL);
 154
 155         if (!mem->va)
 156                 return -ENOMEM;
 157
 158         return 0;
 159 }
 160
 161 /**
 162  * i40e_free_virt_mem_d - OS specific memory free for shared code
 163  * @hw:   pointer to the HW structure
 164  * @mem:  ptr to mem struct to free
 165  **/
 166 int i40e_free_virt_mem_d(struct i40e_hw *hw, struct i40e_virt_mem *mem)
 167 {
 168         /* it's ok to kfree a NULL pointer */
 169         kfree(mem->va);
 170         mem->va = NULL;
 171         mem->size = 0;
 172
 173         return 0;
 174 }
 175
 176 /**
 177  * i40e_get_lump - find a lump of free generic resource
 178  * @pf: board private structure
 179  * @pile: the pile of resource to search
 180  * @needed: the number of items needed
 181  * @id: an owner id to stick on the items assigned
 182  *
 183  * Returns the base item index of the lump, or negative for error
 184  *
 185  * The search_hint trick and lack of advanced fit-finding only work
 186  * because we're highly likely to have all the same size lump requests.
 187  * Linear search time and any fragmentation should be minimal.
 188  **/
 189 static int i40e_get_lump(struct i40e_pf *pf, struct i40e_lump_tracking *pile,
 190                          u16 needed, u16 id)
 191 {
 192         int ret = -ENOMEM;
 193         int i, j;
 194
 195         if (!pile || needed == 0 || id >= I40E_PILE_VALID_BIT) {
 196                 dev_info(&pf->pdev->dev,
 197                          "param err: pile=%p needed=%d id=0x%04x\n",
 198                          pile, needed, id);
 199                 return -EINVAL;
 200         }
 201
 202         /* start the linear search with an imperfect hint */
 203         i = pile->search_hint;
 204         while (i < pile->num_entries) {
 205                 /* skip already allocated entries */
 206                 if (pile->list[i] & I40E_PILE_VALID_BIT) {
 207                         i++;
 208                         continue;
 209                 }
 210
 211                 /* do we have enough in this lump? */
 212                 for (j = 0; (j < needed) && ((i+j) < pile->num_entries); j++) {
 213                         if (pile->list[i+j] & I40E_PILE_VALID_BIT)
 214                                 break;
 215                 }
 216
 217                 if (j == needed) {
 218                         /* there was enough, so assign it to the requestor */
 219                         for (j = 0; j < needed; j++)
 220                                 pile->list[i+j] = id | I40E_PILE_VALID_BIT;
 221                         ret = i;
 222                         pile->search_hint = i + j;
 223                         break;
 224                 }
 225
 226                 /* not enough, so skip over it and continue looking */
 227                 i += j;
 228         }
 229
 230         return ret;
 231 }
 232
 233 /**
 234  * i40e_put_lump - return a lump of generic resource
 235  * @pile: the pile of resource to search
 236  * @index: the base item index
 237  * @id: the owner id of the items assigned
 238  *
 239  * Returns the count of items in the lump
 240  **/
 241 static int i40e_put_lump(struct i40e_lump_tracking *pile, u16 index, u16 id)
 242 {
 243         int valid_id = (id | I40E_PILE_VALID_BIT);
 244         int count = 0;
 245         int i;
 246
 247         if (!pile || index >= pile->num_entries)
 248                 return -EINVAL;
 249
 250         for (i = index;
 251              i < pile->num_entries && pile->list[i] == valid_id;
 252              i++) {
 253                 pile->list[i] = 0;
 254                 count++;
 255         }
 256
 257         if (count && index < pile->search_hint)
 258                 pile->search_hint = index;
 259
 260         return count;
 261 }
 262
 263 /**
 264  * i40e_find_vsi_from_id - searches for the vsi with the given id
 265  * @pf - the pf structure to search for the vsi
 266  * @id - id of the vsi it is searching for
 267  **/
 268 struct i40e_vsi *i40e_find_vsi_from_id(struct i40e_pf *pf, u16 id)
 269 {
 270         int i;
 271
 272         for (i = 0; i < pf->num_alloc_vsi; i++)
 273                 if (pf->vsi[i] && (pf->vsi[i]->id == id))
 274                         return pf->vsi[i];
 275
 276         return NULL;
 277 }
 278
 279 /**
 280  * i40e_service_event_schedule - Schedule the service task to wake up
 281  * @pf: board private structure
 282  *
 283  * If not already scheduled, this puts the task into the work queue
 284  **/
 285 static void i40e_service_event_schedule(struct i40e_pf *pf)
 286 {
 287         if (!test_bit(__I40E_DOWN, &pf->state) &&
 288             !test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state) &&
 289             !test_and_set_bit(__I40E_SERVICE_SCHED, &pf->state))
 290                 schedule_work(&pf->service_task);
 291 }
 292
 293 /**
 294  * i40e_tx_timeout - Respond to a Tx Hang
 295  * @netdev: network interface device structure
 296  *
 297  * If any port has noticed a Tx timeout, it is likely that the whole
 298  * device is munged, not just the one netdev port, so go for the full
 299  * reset.
 300  **/
 301 #ifdef I40E_FCOE
 302 void i40e_tx_timeout(struct net_device *netdev)
 303 #else
 304 static void i40e_tx_timeout(struct net_device *netdev)
 305 #endif
 306 {
 307         struct i40e_netdev_priv *np = netdev_priv(netdev);
 308         struct i40e_vsi *vsi = np->vsi;
 309         struct i40e_pf *pf = vsi->back;
 310         struct i40e_ring *tx_ring = NULL;
 311         unsigned int i, hung_queue = 0;
 312         u32 head, val;
 313
 314         pf->tx_timeout_count++;
 315
 316         /* find the stopped queue the same way the stack does */
 317         for (i = 0; i < netdev->num_tx_queues; i++) {
 318                 struct netdev_queue *q;
 319                 unsigned long trans_start;
 320
 321                 q = netdev_get_tx_queue(netdev, i);
 322                 trans_start = q->trans_start ? : netdev->trans_start;
 323                 if (netif_xmit_stopped(q) &&
 324                     time_after(jiffies,
 325                                (trans_start + netdev->watchdog_timeo))) {
 326                         hung_queue = i;
 327                         break;
 328                 }
 329         }
 330
 331         if (i == netdev->num_tx_queues) {
 332                 netdev_info(netdev, "tx_timeout: no netdev hung queue found\n");
 333         } else {
 334                 /* now that we have an index, find the tx_ring struct */
 335                 for (i = 0; i < vsi->num_queue_pairs; i++) {
 336                         if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc) {
 337                                 if (hung_queue ==
 338                                     vsi->tx_rings[i]->queue_index) {
 339                                         tx_ring = vsi->tx_rings[i];
 340                                         break;
 341                                 }
 342                         }
 343                 }
 344         }
 345
 346         if (time_after(jiffies, (pf->tx_timeout_last_recovery + HZ*20)))
 347                 pf->tx_timeout_recovery_level = 1;  /* reset after some time */
 348         else if (time_before(jiffies,
 349                       (pf->tx_timeout_last_recovery + netdev->watchdog_timeo)))
 350                 return;   /* don't do any new action before the next timeout */
 351
 352         if (tx_ring) {
 353                 head = i40e_get_head(tx_ring);
 354                 /* Read interrupt register */
 355                 if (pf->flags & I40E_FLAG_MSIX_ENABLED)
 356                         val = rd32(&pf->hw,
 357                              I40E_PFINT_DYN_CTLN(tx_ring->q_vector->v_idx +
 358                                                 tx_ring->vsi->base_vector - 1));
 359                 else
 360                         val = rd32(&pf->hw, I40E_PFINT_DYN_CTL0);
 361
 362                 netdev_info(netdev, "tx_timeout: VSI_seid: %d, Q %d, NTC: 0x%x, HWB: 0x%x, NTU: 0x%x, TAIL: 0x%x, INT: 0x%x\n",
 363                             vsi->seid, hung_queue, tx_ring->next_to_clean,
 364                             head, tx_ring->next_to_use,
 365                             readl(tx_ring->tail), val);
 366         }
 367
 368         pf->tx_timeout_last_recovery = jiffies;
 369         netdev_info(netdev, "tx_timeout recovery level %d, hung_queue %d\n",
 370                     pf->tx_timeout_recovery_level, hung_queue);
 371
 372         switch (pf->tx_timeout_recovery_level) {
 373         case 1:
 374                 set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
 375                 break;
 376         case 2:
 377                 set_bit(__I40E_CORE_RESET_REQUESTED, &pf->state);
 378                 break;
 379         case 3:
 380                 set_bit(__I40E_GLOBAL_RESET_REQUESTED, &pf->state);
 381                 break;
 382         default:
 383                 netdev_err(netdev, "tx_timeout recovery unsuccessful\n");
 384                 break;
 385         }
 386
 387         i40e_service_event_schedule(pf);
 388         pf->tx_timeout_recovery_level++;
 389 }
 390
 391 /**
 392  * i40e_release_rx_desc - Store the new tail and head values
 393  * @rx_ring: ring to bump
 394  * @val: new head index
 395  **/
 396 static inline void i40e_release_rx_desc(struct i40e_ring *rx_ring, u32 val)
 397 {
 398         rx_ring->next_to_use = val;
 399
 400         /* Force memory writes to complete before letting h/w
 401          * know there are new descriptors to fetch.  (Only
 402          * applicable for weak-ordered memory model archs,
 403          * such as IA-64).
 404          */
 405         wmb();
 406         writel(val, rx_ring->tail);
 407 }
 408
 409 /**
 410  * i40e_get_vsi_stats_struct - Get System Network Statistics
 411  * @vsi: the VSI we care about
 412  *
 413  * Returns the address of the device statistics structure.
 414  * The statistics are actually updated from the service task.
 415  **/
 416 struct rtnl_link_stats64 *i40e_get_vsi_stats_struct(struct i40e_vsi *vsi)
 417 {
 418         return &vsi->net_stats;
 419 }
 420
 421 /**
 422  * i40e_get_netdev_stats_struct - Get statistics for netdev interface
 423  * @netdev: network interface device structure
 424  *
 425  * Returns the address of the device statistics structure.
 426  * The statistics are actually updated from the service task.
 427  **/
 428 #ifdef I40E_FCOE
 429 struct rtnl_link_stats64 *i40e_get_netdev_stats_struct(
 430                                              struct net_device *netdev,
 431                                              struct rtnl_link_stats64 *stats)
 432 #else
 433 static struct rtnl_link_stats64 *i40e_get_netdev_stats_struct(
 434                                              struct net_device *netdev,
 435                                              struct rtnl_link_stats64 *stats)
 436 #endif
 437 {
 438         struct i40e_netdev_priv *np = netdev_priv(netdev);
 439         struct i40e_ring *tx_ring, *rx_ring;
 440         struct i40e_vsi *vsi = np->vsi;
 441         struct rtnl_link_stats64 *vsi_stats = i40e_get_vsi_stats_struct(vsi);
 442         int i;
 443
 444         if (test_bit(__I40E_DOWN, &vsi->state))
 445                 return stats;
 446
 447         if (!vsi->tx_rings)
 448                 return stats;
 449
 450         rcu_read_lock();
 451         for (i = 0; i < vsi->num_queue_pairs; i++) {
 452                 u64 bytes, packets;
 453                 unsigned int start;
 454
 455                 tx_ring = ACCESS_ONCE(vsi->tx_rings[i]);
 456                 if (!tx_ring)
 457                         continue;
 458
 459                 do {
 460                         start = u64_stats_fetch_begin_irq(&tx_ring->syncp);
 461                         packets = tx_ring->stats.packets;
 462                         bytes   = tx_ring->stats.bytes;
 463                 } while (u64_stats_fetch_retry_irq(&tx_ring->syncp, start));
 464
 465                 stats->tx_packets += packets;
 466                 stats->tx_bytes   += bytes;
 467                 rx_ring = &tx_ring[1];
 468
 469                 do {
 470                         start = u64_stats_fetch_begin_irq(&rx_ring->syncp);
 471                         packets = rx_ring->stats.packets;
 472                         bytes   = rx_ring->stats.bytes;
 473                 } while (u64_stats_fetch_retry_irq(&rx_ring->syncp, start));
 474
 475                 stats->rx_packets += packets;
 476                 stats->rx_bytes   += bytes;
 477         }
 478         rcu_read_unlock();
 479
 480         /* following stats updated by i40e_watchdog_subtask() */
 481         stats->multicast        = vsi_stats->multicast;
 482         stats->tx_errors        = vsi_stats->tx_errors;
 483         stats->tx_dropped       = vsi_stats->tx_dropped;
 484         stats->rx_errors        = vsi_stats->rx_errors;
 485         stats->rx_dropped       = vsi_stats->rx_dropped;
 486         stats->rx_crc_errors    = vsi_stats->rx_crc_errors;
 487         stats->rx_length_errors = vsi_stats->rx_length_errors;
 488
 489         return stats;
 490 }
 491
 492 /**
 493  * i40e_vsi_reset_stats - Resets all stats of the given vsi
 494  * @vsi: the VSI to have its stats reset
 495  **/
 496 void i40e_vsi_reset_stats(struct i40e_vsi *vsi)
 497 {
 498         struct rtnl_link_stats64 *ns;
 499         int i;
 500
 501         if (!vsi)
 502                 return;
 503
 504         ns = i40e_get_vsi_stats_struct(vsi);
 505         memset(ns, 0, sizeof(*ns));
 506         memset(&vsi->net_stats_offsets, 0, sizeof(vsi->net_stats_offsets));
 507         memset(&vsi->eth_stats, 0, sizeof(vsi->eth_stats));
 508         memset(&vsi->eth_stats_offsets, 0, sizeof(vsi->eth_stats_offsets));
 509         if (vsi->rx_rings && vsi->rx_rings[0]) {
 510                 for (i = 0; i < vsi->num_queue_pairs; i++) {
 511                         memset(&vsi->rx_rings[i]->stats, 0,
 512                                sizeof(vsi->rx_rings[i]->stats));
 513                         memset(&vsi->rx_rings[i]->rx_stats, 0,
 514                                sizeof(vsi->rx_rings[i]->rx_stats));
 515                         memset(&vsi->tx_rings[i]->stats, 0,
 516                                sizeof(vsi->tx_rings[i]->stats));
 517                         memset(&vsi->tx_rings[i]->tx_stats, 0,
 518                                sizeof(vsi->tx_rings[i]->tx_stats));
 519                 }
 520         }
 521         vsi->stat_offsets_loaded = false;
 522 }
 523
 524 /**
 525  * i40e_pf_reset_stats - Reset all of the stats for the given PF
 526  * @pf: the PF to be reset
 527  **/
 528 void i40e_pf_reset_stats(struct i40e_pf *pf)
 529 {
 530         int i;
 531
 532         memset(&pf->stats, 0, sizeof(pf->stats));
 533         memset(&pf->stats_offsets, 0, sizeof(pf->stats_offsets));
 534         pf->stat_offsets_loaded = false;
 535
 536         for (i = 0; i < I40E_MAX_VEB; i++) {
 537                 if (pf->veb[i]) {
 538                         memset(&pf->veb[i]->stats, 0,
 539                                sizeof(pf->veb[i]->stats));
 540                         memset(&pf->veb[i]->stats_offsets, 0,
 541                                sizeof(pf->veb[i]->stats_offsets));
 542                         pf->veb[i]->stat_offsets_loaded = false;
 543                 }
 544         }
 545 }
 546
 547 /**
 548  * i40e_stat_update48 - read and update a 48 bit stat from the chip
 549  * @hw: ptr to the hardware info
 550  * @hireg: the high 32 bit reg to read
 551  * @loreg: the low 32 bit reg to read
 552  * @offset_loaded: has the initial offset been loaded yet
 553  * @offset: ptr to current offset value
 554  * @stat: ptr to the stat
 555  *
 556  * Since the device stats are not reset at PFReset, they likely will not
 557  * be zeroed when the driver starts.  We'll save the first values read
 558  * and use them as offsets to be subtracted from the raw values in order
 559  * to report stats that count from zero.  In the process, we also manage
 560  * the potential roll-over.
 561  **/
 562 static void i40e_stat_update48(struct i40e_hw *hw, u32 hireg, u32 loreg,
 563                                bool offset_loaded, u64 *offset, u64 *stat)
 564 {
 565         u64 new_data;
 566
 567         if (hw->device_id == I40E_DEV_ID_QEMU) {
 568                 new_data = rd32(hw, loreg);
 569                 new_data |= ((u64)(rd32(hw, hireg) & 0xFFFF)) << 32;
 570         } else {
 571                 new_data = rd64(hw, loreg);
 572         }
 573         if (!offset_loaded)
 574                 *offset = new_data;
 575         if (likely(new_data >= *offset))
 576                 *stat = new_data - *offset;
 577         else
 578                 *stat = (new_data + BIT_ULL(48)) - *offset;
 579         *stat &= 0xFFFFFFFFFFFFULL;
 580 }
 581
 582 /**
 583  * i40e_stat_update32 - read and update a 32 bit stat from the chip
 584  * @hw: ptr to the hardware info
 585  * @reg: the hw reg to read
 586  * @offset_loaded: has the initial offset been loaded yet
 587  * @offset: ptr to current offset value
 588  * @stat: ptr to the stat
 589  **/
 590 static void i40e_stat_update32(struct i40e_hw *hw, u32 reg,
 591                                bool offset_loaded, u64 *offset, u64 *stat)
 592 {
 593         u32 new_data;
 594
 595         new_data = rd32(hw, reg);
 596         if (!offset_loaded)
 597                 *offset = new_data;
 598         if (likely(new_data >= *offset))
 599                 *stat = (u32)(new_data - *offset);
 600         else
 601                 *stat = (u32)((new_data + BIT_ULL(32)) - *offset);
 602 }
 603
 604 /**
 605  * i40e_update_eth_stats - Update VSI-specific ethernet statistics counters.
 606  * @vsi: the VSI to be updated
 607  **/
 608 void i40e_update_eth_stats(struct i40e_vsi *vsi)
 609 {
 610         int stat_idx = le16_to_cpu(vsi->info.stat_counter_idx);
 611         struct i40e_pf *pf = vsi->back;
 612         struct i40e_hw *hw = &pf->hw;
 613         struct i40e_eth_stats *oes;
 614         struct i40e_eth_stats *es;     /* device's eth stats */
 615
 616         es = &vsi->eth_stats;
 617         oes = &vsi->eth_stats_offsets;
 618
 619         /* Gather up the stats that the hw collects */
 620         i40e_stat_update32(hw, I40E_GLV_TEPC(stat_idx),
 621                            vsi->stat_offsets_loaded,
 622                            &oes->tx_errors, &es->tx_errors);
 623         i40e_stat_update32(hw, I40E_GLV_RDPC(stat_idx),
 624                            vsi->stat_offsets_loaded,
 625                            &oes->rx_discards, &es->rx_discards);
 626         i40e_stat_update32(hw, I40E_GLV_RUPP(stat_idx),
 627                            vsi->stat_offsets_loaded,
 628                            &oes->rx_unknown_protocol, &es->rx_unknown_protocol);
 629         i40e_stat_update32(hw, I40E_GLV_TEPC(stat_idx),
 630                            vsi->stat_offsets_loaded,
 631                            &oes->tx_errors, &es->tx_errors);
 632
 633         i40e_stat_update48(hw, I40E_GLV_GORCH(stat_idx),
 634                            I40E_GLV_GORCL(stat_idx),
 635                            vsi->stat_offsets_loaded,
 636                            &oes->rx_bytes, &es->rx_bytes);
 637         i40e_stat_update48(hw, I40E_GLV_UPRCH(stat_idx),
 638                            I40E_GLV_UPRCL(stat_idx),
 639                            vsi->stat_offsets_loaded,
 640                            &oes->rx_unicast, &es->rx_unicast);
 641         i40e_stat_update48(hw, I40E_GLV_MPRCH(stat_idx),
 642                            I40E_GLV_MPRCL(stat_idx),
 643                            vsi->stat_offsets_loaded,
 644                            &oes->rx_multicast, &es->rx_multicast);
 645         i40e_stat_update48(hw, I40E_GLV_BPRCH(stat_idx),
 646                            I40E_GLV_BPRCL(stat_idx),
 647                            vsi->stat_offsets_loaded,
 648                            &oes->rx_broadcast, &es->rx_broadcast);
 649
 650         i40e_stat_update48(hw, I40E_GLV_GOTCH(stat_idx),
 651                            I40E_GLV_GOTCL(stat_idx),
 652                            vsi->stat_offsets_loaded,
 653                            &oes->tx_bytes, &es->tx_bytes);
 654         i40e_stat_update48(hw, I40E_GLV_UPTCH(stat_idx),
 655                            I40E_GLV_UPTCL(stat_idx),
 656                            vsi->stat_offsets_loaded,
 657                            &oes->tx_unicast, &es->tx_unicast);
 658         i40e_stat_update48(hw, I40E_GLV_MPTCH(stat_idx),
 659                            I40E_GLV_MPTCL(stat_idx),
 660                            vsi->stat_offsets_loaded,
 661                            &oes->tx_multicast, &es->tx_multicast);
 662         i40e_stat_update48(hw, I40E_GLV_BPTCH(stat_idx),
 663                            I40E_GLV_BPTCL(stat_idx),
 664                            vsi->stat_offsets_loaded,
 665                            &oes->tx_broadcast, &es->tx_broadcast);
 666         vsi->stat_offsets_loaded = true;
 667 }
 668
 669 /**
 670  * i40e_update_veb_stats - Update Switch component statistics
 671  * @veb: the VEB being updated
 672  **/
 673 static void i40e_update_veb_stats(struct i40e_veb *veb)
 674 {
 675         struct i40e_pf *pf = veb->pf;
 676         struct i40e_hw *hw = &pf->hw;
 677         struct i40e_eth_stats *oes;
 678         struct i40e_eth_stats *es;     /* device's eth stats */
 679         struct i40e_veb_tc_stats *veb_oes;
 680         struct i40e_veb_tc_stats *veb_es;
 681         int i, idx = 0;
 682
 683         idx = veb->stats_idx;
 684         es = &veb->stats;
 685         oes = &veb->stats_offsets;
 686         veb_es = &veb->tc_stats;
 687         veb_oes = &veb->tc_stats_offsets;
 688
 689         /* Gather up the stats that the hw collects */
 690         i40e_stat_update32(hw, I40E_GLSW_TDPC(idx),
 691                            veb->stat_offsets_loaded,
 692                            &oes->tx_discards, &es->tx_discards);
 693         if (hw->revision_id > 0)
 694                 i40e_stat_update32(hw, I40E_GLSW_RUPP(idx),
 695                                    veb->stat_offsets_loaded,
 696                                    &oes->rx_unknown_protocol,
 697                                    &es->rx_unknown_protocol);
 698         i40e_stat_update48(hw, I40E_GLSW_GORCH(idx), I40E_GLSW_GORCL(idx),
 699                            veb->stat_offsets_loaded,
 700                            &oes->rx_bytes, &es->rx_bytes);
 701         i40e_stat_update48(hw, I40E_GLSW_UPRCH(idx), I40E_GLSW_UPRCL(idx),
 702                            veb->stat_offsets_loaded,
 703                            &oes->rx_unicast, &es->rx_unicast);
 704         i40e_stat_update48(hw, I40E_GLSW_MPRCH(idx), I40E_GLSW_MPRCL(idx),
 705                            veb->stat_offsets_loaded,
 706                            &oes->rx_multicast, &es->rx_multicast);
 707         i40e_stat_update48(hw, I40E_GLSW_BPRCH(idx), I40E_GLSW_BPRCL(idx),
 708                            veb->stat_offsets_loaded,
 709                            &oes->rx_broadcast, &es->rx_broadcast);
 710
 711         i40e_stat_update48(hw, I40E_GLSW_GOTCH(idx), I40E_GLSW_GOTCL(idx),
 712                            veb->stat_offsets_loaded,
 713                            &oes->tx_bytes, &es->tx_bytes);
 714         i40e_stat_update48(hw, I40E_GLSW_UPTCH(idx), I40E_GLSW_UPTCL(idx),
 715                            veb->stat_offsets_loaded,
 716                            &oes->tx_unicast, &es->tx_unicast);
 717         i40e_stat_update48(hw, I40E_GLSW_MPTCH(idx), I40E_GLSW_MPTCL(idx),
 718                            veb->stat_offsets_loaded,
 719                            &oes->tx_multicast, &es->tx_multicast);
 720         i40e_stat_update48(hw, I40E_GLSW_BPTCH(idx), I40E_GLSW_BPTCL(idx),
 721                            veb->stat_offsets_loaded,
 722                            &oes->tx_broadcast, &es->tx_broadcast);
 723         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
 724                 i40e_stat_update48(hw, I40E_GLVEBTC_RPCH(i, idx),
 725                                    I40E_GLVEBTC_RPCL(i, idx),
 726                                    veb->stat_offsets_loaded,
 727                                    &veb_oes->tc_rx_packets[i],
 728                                    &veb_es->tc_rx_packets[i]);
 729                 i40e_stat_update48(hw, I40E_GLVEBTC_RBCH(i, idx),
 730                                    I40E_GLVEBTC_RBCL(i, idx),
 731                                    veb->stat_offsets_loaded,
 732                                    &veb_oes->tc_rx_bytes[i],
 733                                    &veb_es->tc_rx_bytes[i]);
 734                 i40e_stat_update48(hw, I40E_GLVEBTC_TPCH(i, idx),
 735                                    I40E_GLVEBTC_TPCL(i, idx),
 736                                    veb->stat_offsets_loaded,
 737                                    &veb_oes->tc_tx_packets[i],
 738                                    &veb_es->tc_tx_packets[i]);
 739                 i40e_stat_update48(hw, I40E_GLVEBTC_TBCH(i, idx),
 740                                    I40E_GLVEBTC_TBCL(i, idx),
 741                                    veb->stat_offsets_loaded,
 742                                    &veb_oes->tc_tx_bytes[i],
 743                                    &veb_es->tc_tx_bytes[i]);
 744         }
 745         veb->stat_offsets_loaded = true;
 746 }
 747
 748 #ifdef I40E_FCOE
 749 /**
 750  * i40e_update_fcoe_stats - Update FCoE-specific ethernet statistics counters.
 751  * @vsi: the VSI that is capable of doing FCoE
 752  **/
 753 static void i40e_update_fcoe_stats(struct i40e_vsi *vsi)
 754 {
 755         struct i40e_pf *pf = vsi->back;
 756         struct i40e_hw *hw = &pf->hw;
 757         struct i40e_fcoe_stats *ofs;
 758         struct i40e_fcoe_stats *fs;     /* device's eth stats */
 759         int idx;
 760
 761         if (vsi->type != I40E_VSI_FCOE)
 762                 return;
 763
 764         idx = (pf->pf_seid - I40E_BASE_PF_SEID) + I40E_FCOE_PF_STAT_OFFSET;
 765         fs = &vsi->fcoe_stats;
 766         ofs = &vsi->fcoe_stats_offsets;
 767
 768         i40e_stat_update32(hw, I40E_GL_FCOEPRC(idx),
 769                            vsi->fcoe_stat_offsets_loaded,
 770                            &ofs->rx_fcoe_packets, &fs->rx_fcoe_packets);
 771         i40e_stat_update48(hw, I40E_GL_FCOEDWRCH(idx), I40E_GL_FCOEDWRCL(idx),
 772                            vsi->fcoe_stat_offsets_loaded,
 773                            &ofs->rx_fcoe_dwords, &fs->rx_fcoe_dwords);
 774         i40e_stat_update32(hw, I40E_GL_FCOERPDC(idx),
 775                            vsi->fcoe_stat_offsets_loaded,
 776                            &ofs->rx_fcoe_dropped, &fs->rx_fcoe_dropped);
 777         i40e_stat_update32(hw, I40E_GL_FCOEPTC(idx),
 778                            vsi->fcoe_stat_offsets_loaded,
 779                            &ofs->tx_fcoe_packets, &fs->tx_fcoe_packets);
 780         i40e_stat_update48(hw, I40E_GL_FCOEDWTCH(idx), I40E_GL_FCOEDWTCL(idx),
 781                            vsi->fcoe_stat_offsets_loaded,
 782                            &ofs->tx_fcoe_dwords, &fs->tx_fcoe_dwords);
 783         i40e_stat_update32(hw, I40E_GL_FCOECRC(idx),
 784                            vsi->fcoe_stat_offsets_loaded,
 785                            &ofs->fcoe_bad_fccrc, &fs->fcoe_bad_fccrc);
 786         i40e_stat_update32(hw, I40E_GL_FCOELAST(idx),
 787                            vsi->fcoe_stat_offsets_loaded,
 788                            &ofs->fcoe_last_error, &fs->fcoe_last_error);
 789         i40e_stat_update32(hw, I40E_GL_FCOEDDPC(idx),
 790                            vsi->fcoe_stat_offsets_loaded,
 791                            &ofs->fcoe_ddp_count, &fs->fcoe_ddp_count);
 792
 793         vsi->fcoe_stat_offsets_loaded = true;
 794 }
 795
 796 #endif
 797 /**
 798  * i40e_update_vsi_stats - Update the vsi statistics counters.
 799  * @vsi: the VSI to be updated
 800  *
 801  * There are a few instances where we store the same stat in a
 802  * couple of different structs.  This is partly because we have
 803  * the netdev stats that need to be filled out, which is slightly
 804  * different from the "eth_stats" defined by the chip and used in
 805  * VF communications.  We sort it out here.
 806  **/
 807 static void i40e_update_vsi_stats(struct i40e_vsi *vsi)
 808 {
 809         struct i40e_pf *pf = vsi->back;
 810         struct rtnl_link_stats64 *ons;
 811         struct rtnl_link_stats64 *ns;   /* netdev stats */
 812         struct i40e_eth_stats *oes;
 813         struct i40e_eth_stats *es;     /* device's eth stats */
 814         u32 tx_restart, tx_busy;
 815         struct i40e_ring *p;
 816         u32 rx_page, rx_buf;
 817         u64 bytes, packets;
 818         unsigned int start;
 819         u64 tx_linearize;
 820         u64 tx_force_wb;
 821         u64 rx_p, rx_b;
 822         u64 tx_p, tx_b;
 823         u16 q;
 824
 825         if (test_bit(__I40E_DOWN, &vsi->state) ||
 826             test_bit(__I40E_CONFIG_BUSY, &pf->state))
 827                 return;
 828
 829         ns = i40e_get_vsi_stats_struct(vsi);
 830         ons = &vsi->net_stats_offsets;
 831         es = &vsi->eth_stats;
 832         oes = &vsi->eth_stats_offsets;
 833
 834         /* Gather up the netdev and vsi stats that the driver collects
 835          * on the fly during packet processing
 836          */
 837         rx_b = rx_p = 0;
 838         tx_b = tx_p = 0;
 839         tx_restart = tx_busy = tx_linearize = tx_force_wb = 0;
 840         rx_page = 0;
 841         rx_buf = 0;
 842         rcu_read_lock();
 843         for (q = 0; q < vsi->num_queue_pairs; q++) {
 844                 /* locate Tx ring */
 845                 p = ACCESS_ONCE(vsi->tx_rings[q]);
 846
 847                 do {
 848                         start = u64_stats_fetch_begin_irq(&p->syncp);
 849                         packets = p->stats.packets;
 850                         bytes = p->stats.bytes;
 851                 } while (u64_stats_fetch_retry_irq(&p->syncp, start));
 852                 tx_b += bytes;
 853                 tx_p += packets;
 854                 tx_restart += p->tx_stats.restart_queue;
 855                 tx_busy += p->tx_stats.tx_busy;
 856                 tx_linearize += p->tx_stats.tx_linearize;
 857                 tx_force_wb += p->tx_stats.tx_force_wb;
 858
 859                 /* Rx queue is part of the same block as Tx queue */
 860                 p = &p[1];
 861                 do {
 862                         start = u64_stats_fetch_begin_irq(&p->syncp);
 863                         packets = p->stats.packets;
 864                         bytes = p->stats.bytes;
 865                 } while (u64_stats_fetch_retry_irq(&p->syncp, start));
 866                 rx_b += bytes;
 867                 rx_p += packets;
 868                 rx_buf += p->rx_stats.alloc_buff_failed;
 869                 rx_page += p->rx_stats.alloc_page_failed;
 870         }
 871         rcu_read_unlock();
 872         vsi->tx_restart = tx_restart;
 873         vsi->tx_busy = tx_busy;
 874         vsi->tx_linearize = tx_linearize;
 875         vsi->tx_force_wb = tx_force_wb;
 876         vsi->rx_page_failed = rx_page;
 877         vsi->rx_buf_failed = rx_buf;
 878
 879         ns->rx_packets = rx_p;
 880         ns->rx_bytes = rx_b;
 881         ns->tx_packets = tx_p;
 882         ns->tx_bytes = tx_b;
 883
 884         /* update netdev stats from eth stats */
 885         i40e_update_eth_stats(vsi);
 886         ons->tx_errors = oes->tx_errors;
 887         ns->tx_errors = es->tx_errors;
 888         ons->multicast = oes->rx_multicast;
 889         ns->multicast = es->rx_multicast;
 890         ons->rx_dropped = oes->rx_discards;
 891         ns->rx_dropped = es->rx_discards;
 892         ons->tx_dropped = oes->tx_discards;
 893         ns->tx_dropped = es->tx_discards;
 894
 895         /* pull in a couple PF stats if this is the main vsi */
 896         if (vsi == pf->vsi[pf->lan_vsi]) {
 897                 ns->rx_crc_errors = pf->stats.crc_errors;
 898                 ns->rx_errors = pf->stats.crc_errors + pf->stats.illegal_bytes;
 899                 ns->rx_length_errors = pf->stats.rx_length_errors;
 900         }
 901 }
 902
 903 /**
 904  * i40e_update_pf_stats - Update the PF statistics counters.
 905  * @pf: the PF to be updated
 906  **/
 907 static void i40e_update_pf_stats(struct i40e_pf *pf)
 908 {
 909         struct i40e_hw_port_stats *osd = &pf->stats_offsets;
 910         struct i40e_hw_port_stats *nsd = &pf->stats;
 911         struct i40e_hw *hw = &pf->hw;
 912         u32 val;
 913         int i;
 914
 915         i40e_stat_update48(hw, I40E_GLPRT_GORCH(hw->port),
 916                            I40E_GLPRT_GORCL(hw->port),
 917                            pf->stat_offsets_loaded,
 918                            &osd->eth.rx_bytes, &nsd->eth.rx_bytes);
 919         i40e_stat_update48(hw, I40E_GLPRT_GOTCH(hw->port),
 920                            I40E_GLPRT_GOTCL(hw->port),
 921                            pf->stat_offsets_loaded,
 922                            &osd->eth.tx_bytes, &nsd->eth.tx_bytes);
 923         i40e_stat_update32(hw, I40E_GLPRT_RDPC(hw->port),
 924                            pf->stat_offsets_loaded,
 925                            &osd->eth.rx_discards,
 926                            &nsd->eth.rx_discards);
 927         i40e_stat_update48(hw, I40E_GLPRT_UPRCH(hw->port),
 928                            I40E_GLPRT_UPRCL(hw->port),
 929                            pf->stat_offsets_loaded,
 930                            &osd->eth.rx_unicast,
 931                            &nsd->eth.rx_unicast);
 932         i40e_stat_update48(hw, I40E_GLPRT_MPRCH(hw->port),
 933                            I40E_GLPRT_MPRCL(hw->port),
 934                            pf->stat_offsets_loaded,
 935                            &osd->eth.rx_multicast,
 936                            &nsd->eth.rx_multicast);
 937         i40e_stat_update48(hw, I40E_GLPRT_BPRCH(hw->port),
 938                            I40E_GLPRT_BPRCL(hw->port),
 939                            pf->stat_offsets_loaded,
 940                            &osd->eth.rx_broadcast,
 941                            &nsd->eth.rx_broadcast);
 942         i40e_stat_update48(hw, I40E_GLPRT_UPTCH(hw->port),
 943                            I40E_GLPRT_UPTCL(hw->port),
 944                            pf->stat_offsets_loaded,
 945                            &osd->eth.tx_unicast,
 946                            &nsd->eth.tx_unicast);
 947         i40e_stat_update48(hw, I40E_GLPRT_MPTCH(hw->port),
 948                            I40E_GLPRT_MPTCL(hw->port),
 949                            pf->stat_offsets_loaded,
 950                            &osd->eth.tx_multicast,
 951                            &nsd->eth.tx_multicast);
 952         i40e_stat_update48(hw, I40E_GLPRT_BPTCH(hw->port),
 953                            I40E_GLPRT_BPTCL(hw->port),
 954                            pf->stat_offsets_loaded,
 955                            &osd->eth.tx_broadcast,
 956                            &nsd->eth.tx_broadcast);
 957
 958         i40e_stat_update32(hw, I40E_GLPRT_TDOLD(hw->port),
 959                            pf->stat_offsets_loaded,
 960                            &osd->tx_dropped_link_down,
 961                            &nsd->tx_dropped_link_down);
 962
 963         i40e_stat_update32(hw, I40E_GLPRT_CRCERRS(hw->port),
 964                            pf->stat_offsets_loaded,
 965                            &osd->crc_errors, &nsd->crc_errors);
 966
 967         i40e_stat_update32(hw, I40E_GLPRT_ILLERRC(hw->port),
 968                            pf->stat_offsets_loaded,
 969                            &osd->illegal_bytes, &nsd->illegal_bytes);
 970
 971         i40e_stat_update32(hw, I40E_GLPRT_MLFC(hw->port),
 972                            pf->stat_offsets_loaded,
 973                            &osd->mac_local_faults,
 974                            &nsd->mac_local_faults);
 975         i40e_stat_update32(hw, I40E_GLPRT_MRFC(hw->port),
 976                            pf->stat_offsets_loaded,
 977                            &osd->mac_remote_faults,
 978                            &nsd->mac_remote_faults);
 979
 980         i40e_stat_update32(hw, I40E_GLPRT_RLEC(hw->port),
 981                            pf->stat_offsets_loaded,
 982                            &osd->rx_length_errors,
 983                            &nsd->rx_length_errors);
 984
 985         i40e_stat_update32(hw, I40E_GLPRT_LXONRXC(hw->port),
 986                            pf->stat_offsets_loaded,
 987                            &osd->link_xon_rx, &nsd->link_xon_rx);
 988         i40e_stat_update32(hw, I40E_GLPRT_LXONTXC(hw->port),
 989                            pf->stat_offsets_loaded,
 990                            &osd->link_xon_tx, &nsd->link_xon_tx);
 991         i40e_stat_update32(hw, I40E_GLPRT_LXOFFRXC(hw->port),
 992                            pf->stat_offsets_loaded,
 993                            &osd->link_xoff_rx, &nsd->link_xoff_rx);
 994         i40e_stat_update32(hw, I40E_GLPRT_LXOFFTXC(hw->port),
 995                            pf->stat_offsets_loaded,
 996                            &osd->link_xoff_tx, &nsd->link_xoff_tx);
 997
 998         for (i = 0; i < 8; i++) {
 999                 i40e_stat_update32(hw, I40E_GLPRT_PXOFFRXC(hw->port, i),
1000                                    pf->stat_offsets_loaded,
1001                                    &osd->priority_xoff_rx[i],
1002                                    &nsd->priority_xoff_rx[i]);
1003                 i40e_stat_update32(hw, I40E_GLPRT_PXONRXC(hw->port, i),
1004                                    pf->stat_offsets_loaded,
1005                                    &osd->priority_xon_rx[i],
1006                                    &nsd->priority_xon_rx[i]);
1007                 i40e_stat_update32(hw, I40E_GLPRT_PXONTXC(hw->port, i),
1008                                    pf->stat_offsets_loaded,
1009                                    &osd->priority_xon_tx[i],
1010                                    &nsd->priority_xon_tx[i]);
1011                 i40e_stat_update32(hw, I40E_GLPRT_PXOFFTXC(hw->port, i),
1012                                    pf->stat_offsets_loaded,
1013                                    &osd->priority_xoff_tx[i],
1014                                    &nsd->priority_xoff_tx[i]);
1015                 i40e_stat_update32(hw,
1016                                    I40E_GLPRT_RXON2OFFCNT(hw->port, i),
1017                                    pf->stat_offsets_loaded,
1018                                    &osd->priority_xon_2_xoff[i],
1019                                    &nsd->priority_xon_2_xoff[i]);
1020         }
1021
1022         i40e_stat_update48(hw, I40E_GLPRT_PRC64H(hw->port),
1023                            I40E_GLPRT_PRC64L(hw->port),
1024                            pf->stat_offsets_loaded,
1025                            &osd->rx_size_64, &nsd->rx_size_64);
1026         i40e_stat_update48(hw, I40E_GLPRT_PRC127H(hw->port),
1027                            I40E_GLPRT_PRC127L(hw->port),
1028                            pf->stat_offsets_loaded,
1029                            &osd->rx_size_127, &nsd->rx_size_127);
1030         i40e_stat_update48(hw, I40E_GLPRT_PRC255H(hw->port),
1031                            I40E_GLPRT_PRC255L(hw->port),
1032                            pf->stat_offsets_loaded,
1033                            &osd->rx_size_255, &nsd->rx_size_255);
1034         i40e_stat_update48(hw, I40E_GLPRT_PRC511H(hw->port),
1035                            I40E_GLPRT_PRC511L(hw->port),
1036                            pf->stat_offsets_loaded,
1037                            &osd->rx_size_511, &nsd->rx_size_511);
1038         i40e_stat_update48(hw, I40E_GLPRT_PRC1023H(hw->port),
1039                            I40E_GLPRT_PRC1023L(hw->port),
1040                            pf->stat_offsets_loaded,
1041                            &osd->rx_size_1023, &nsd->rx_size_1023);
1042         i40e_stat_update48(hw, I40E_GLPRT_PRC1522H(hw->port),
1043                            I40E_GLPRT_PRC1522L(hw->port),
1044                            pf->stat_offsets_loaded,
1045                            &osd->rx_size_1522, &nsd->rx_size_1522);
1046         i40e_stat_update48(hw, I40E_GLPRT_PRC9522H(hw->port),
1047                            I40E_GLPRT_PRC9522L(hw->port),
1048                            pf->stat_offsets_loaded,
1049                            &osd->rx_size_big, &nsd->rx_size_big);
1050
1051         i40e_stat_update48(hw, I40E_GLPRT_PTC64H(hw->port),
1052                            I40E_GLPRT_PTC64L(hw->port),
1053                            pf->stat_offsets_loaded,
1054                            &osd->tx_size_64, &nsd->tx_size_64);
1055         i40e_stat_update48(hw, I40E_GLPRT_PTC127H(hw->port),
1056                            I40E_GLPRT_PTC127L(hw->port),
1057                            pf->stat_offsets_loaded,
1058                            &osd->tx_size_127, &nsd->tx_size_127);
1059         i40e_stat_update48(hw, I40E_GLPRT_PTC255H(hw->port),
1060                            I40E_GLPRT_PTC255L(hw->port),
1061                            pf->stat_offsets_loaded,
1062                            &osd->tx_size_255, &nsd->tx_size_255);
1063         i40e_stat_update48(hw, I40E_GLPRT_PTC511H(hw->port),
1064                            I40E_GLPRT_PTC511L(hw->port),
1065                            pf->stat_offsets_loaded,
1066                            &osd->tx_size_511, &nsd->tx_size_511);
1067         i40e_stat_update48(hw, I40E_GLPRT_PTC1023H(hw->port),
1068                            I40E_GLPRT_PTC1023L(hw->port),
1069                            pf->stat_offsets_loaded,
1070                            &osd->tx_size_1023, &nsd->tx_size_1023);
1071         i40e_stat_update48(hw, I40E_GLPRT_PTC1522H(hw->port),
1072                            I40E_GLPRT_PTC1522L(hw->port),
1073                            pf->stat_offsets_loaded,
1074                            &osd->tx_size_1522, &nsd->tx_size_1522);
1075         i40e_stat_update48(hw, I40E_GLPRT_PTC9522H(hw->port),
1076                            I40E_GLPRT_PTC9522L(hw->port),
1077                            pf->stat_offsets_loaded,
1078                            &osd->tx_size_big, &nsd->tx_size_big);
1079
1080         i40e_stat_update32(hw, I40E_GLPRT_RUC(hw->port),
1081                            pf->stat_offsets_loaded,
1082                            &osd->rx_undersize, &nsd->rx_undersize);
1083         i40e_stat_update32(hw, I40E_GLPRT_RFC(hw->port),
1084                            pf->stat_offsets_loaded,
1085                            &osd->rx_fragments, &nsd->rx_fragments);
1086         i40e_stat_update32(hw, I40E_GLPRT_ROC(hw->port),
1087                            pf->stat_offsets_loaded,
1088                            &osd->rx_oversize, &nsd->rx_oversize);
1089         i40e_stat_update32(hw, I40E_GLPRT_RJC(hw->port),
1090                            pf->stat_offsets_loaded,
1091                            &osd->rx_jabber, &nsd->rx_jabber);
1092
1093         /* FDIR stats */
1094         i40e_stat_update32(hw,
1095                            I40E_GLQF_PCNT(I40E_FD_ATR_STAT_IDX(pf->hw.pf_id)),
1096                            pf->stat_offsets_loaded,
1097                            &osd->fd_atr_match, &nsd->fd_atr_match);
1098         i40e_stat_update32(hw,
1099                            I40E_GLQF_PCNT(I40E_FD_SB_STAT_IDX(pf->hw.pf_id)),
1100                            pf->stat_offsets_loaded,
1101                            &osd->fd_sb_match, &nsd->fd_sb_match);
1102         i40e_stat_update32(hw,
1103                       I40E_GLQF_PCNT(I40E_FD_ATR_TUNNEL_STAT_IDX(pf->hw.pf_id)),
1104                       pf->stat_offsets_loaded,
1105                       &osd->fd_atr_tunnel_match, &nsd->fd_atr_tunnel_match);
1106
1107         val = rd32(hw, I40E_PRTPM_EEE_STAT);
1108         nsd->tx_lpi_status =
1109                        (val & I40E_PRTPM_EEE_STAT_TX_LPI_STATUS_MASK) >>
1110                         I40E_PRTPM_EEE_STAT_TX_LPI_STATUS_SHIFT;
1111         nsd->rx_lpi_status =
1112                        (val & I40E_PRTPM_EEE_STAT_RX_LPI_STATUS_MASK) >>
1113                         I40E_PRTPM_EEE_STAT_RX_LPI_STATUS_SHIFT;
1114         i40e_stat_update32(hw, I40E_PRTPM_TLPIC,
1115                            pf->stat_offsets_loaded,
1116                            &osd->tx_lpi_count, &nsd->tx_lpi_count);
1117         i40e_stat_update32(hw, I40E_PRTPM_RLPIC,
1118                            pf->stat_offsets_loaded,
1119                            &osd->rx_lpi_count, &nsd->rx_lpi_count);
1120
1121         if (pf->flags & I40E_FLAG_FD_SB_ENABLED &&
1122             !(pf->auto_disable_flags & I40E_FLAG_FD_SB_ENABLED))
1123                 nsd->fd_sb_status = true;
1124         else
1125                 nsd->fd_sb_status = false;
1126
1127         if (pf->flags & I40E_FLAG_FD_ATR_ENABLED &&
1128             !(pf->auto_disable_flags & I40E_FLAG_FD_ATR_ENABLED))
1129                 nsd->fd_atr_status = true;
1130         else
1131                 nsd->fd_atr_status = false;
1132
1133         pf->stat_offsets_loaded = true;
1134 }
1135
1136 /**
1137  * i40e_update_stats - Update the various statistics counters.
1138  * @vsi: the VSI to be updated
1139  *
1140  * Update the various stats for this VSI and its related entities.
1141  **/
1142 void i40e_update_stats(struct i40e_vsi *vsi)
1143 {
1144         struct i40e_pf *pf = vsi->back;
1145
1146         if (vsi == pf->vsi[pf->lan_vsi])
1147                 i40e_update_pf_stats(pf);
1148
1149         i40e_update_vsi_stats(vsi);
1150 #ifdef I40E_FCOE
1151         i40e_update_fcoe_stats(vsi);
1152 #endif
1153 }
1154
1155 /**
1156  * i40e_find_filter - Search VSI filter list for specific mac/vlan filter
1157  * @vsi: the VSI to be searched
1158  * @macaddr: the MAC address
1159  * @vlan: the vlan
1160  * @is_vf: make sure its a VF filter, else doesn't matter
1161  * @is_netdev: make sure its a netdev filter, else doesn't matter
1162  *
1163  * Returns ptr to the filter object or NULL
1164  **/
1165 static struct i40e_mac_filter *i40e_find_filter(struct i40e_vsi *vsi,
1166                                                 u8 *macaddr, s16 vlan,
1167                                                 bool is_vf, bool is_netdev)
1168 {
1169         struct i40e_mac_filter *f;
1170
1171         if (!vsi || !macaddr)
1172                 return NULL;
1173
1174         list_for_each_entry(f, &vsi->mac_filter_list, list) {
1175                 if ((ether_addr_equal(macaddr, f->macaddr)) &&
1176                     (vlan == f->vlan)    &&
1177                     (!is_vf || f->is_vf) &&
1178                     (!is_netdev || f->is_netdev))
1179                         return f;
1180         }
1181         return NULL;
1182 }
1183
1184 /**
1185  * i40e_find_mac - Find a mac addr in the macvlan filters list
1186  * @vsi: the VSI to be searched
1187  * @macaddr: the MAC address we are searching for
1188  * @is_vf: make sure its a VF filter, else doesn't matter
1189  * @is_netdev: make sure its a netdev filter, else doesn't matter
1190  *
1191  * Returns the first filter with the provided MAC address or NULL if
1192  * MAC address was not found
1193  **/
1194 struct i40e_mac_filter *i40e_find_mac(struct i40e_vsi *vsi, u8 *macaddr,
1195                                       bool is_vf, bool is_netdev)
1196 {
1197         struct i40e_mac_filter *f;
1198
1199         if (!vsi || !macaddr)
1200                 return NULL;
1201
1202         list_for_each_entry(f, &vsi->mac_filter_list, list) {
1203                 if ((ether_addr_equal(macaddr, f->macaddr)) &&
1204                     (!is_vf || f->is_vf) &&
1205                     (!is_netdev || f->is_netdev))
1206                         return f;
1207         }
1208         return NULL;
1209 }
1210
1211 /**
1212  * i40e_is_vsi_in_vlan - Check if VSI is in vlan mode
1213  * @vsi: the VSI to be searched
1214  *
1215  * Returns true if VSI is in vlan mode or false otherwise
1216  **/
1217 bool i40e_is_vsi_in_vlan(struct i40e_vsi *vsi)
1218 {
1219         struct i40e_mac_filter *f;
1220
1221         /* Only -1 for all the filters denotes not in vlan mode
1222          * so we have to go through all the list in order to make sure
1223          */
1224         list_for_each_entry(f, &vsi->mac_filter_list, list) {
1225                 if (f->vlan >= 0 || vsi->info.pvid)
1226                         return true;
1227         }
1228
1229         return false;
1230 }
1231
1232 /**
1233  * i40e_put_mac_in_vlan - Make macvlan filters from macaddrs and vlans
1234  * @vsi: the VSI to be searched
1235  * @macaddr: the mac address to be filtered
1236  * @is_vf: true if it is a VF
1237  * @is_netdev: true if it is a netdev
1238  *
1239  * Goes through all the macvlan filters and adds a
1240  * macvlan filter for each unique vlan that already exists
1241  *
1242  * Returns first filter found on success, else NULL
1243  **/
1244 struct i40e_mac_filter *i40e_put_mac_in_vlan(struct i40e_vsi *vsi, u8 *macaddr,
1245                                              bool is_vf, bool is_netdev)
1246 {
1247         struct i40e_mac_filter *f;
1248
1249         list_for_each_entry(f, &vsi->mac_filter_list, list) {
1250                 if (vsi->info.pvid)
1251                         f->vlan = le16_to_cpu(vsi->info.pvid);
1252                 if (!i40e_find_filter(vsi, macaddr, f->vlan,
1253                                       is_vf, is_netdev)) {
1254                         if (!i40e_add_filter(vsi, macaddr, f->vlan,
1255                                              is_vf, is_netdev))
1256                                 return NULL;
1257                 }
1258         }
1259
1260         return list_first_entry_or_null(&vsi->mac_filter_list,
1261                                         struct i40e_mac_filter, list);
1262 }
1263
1264 /**
1265  * i40e_del_mac_all_vlan - Remove a MAC filter from all VLANS
1266  * @vsi: the VSI to be searched
1267  * @macaddr: the mac address to be removed
1268  * @is_vf: true if it is a VF
1269  * @is_netdev: true if it is a netdev
1270  *
1271  * Removes a given MAC address from a VSI, regardless of VLAN
1272  *
1273  * Returns 0 for success, or error
1274  **/
1275 int i40e_del_mac_all_vlan(struct i40e_vsi *vsi, u8 *macaddr,
1276                           bool is_vf, bool is_netdev)
1277 {
1278         struct i40e_mac_filter *f = NULL;
1279         int changed = 0;
1280
1281         WARN(!spin_is_locked(&vsi->mac_filter_list_lock),
1282              "Missing mac_filter_list_lock\n");
1283         list_for_each_entry(f, &vsi->mac_filter_list, list) {
1284                 if ((ether_addr_equal(macaddr, f->macaddr)) &&
1285                     (is_vf == f->is_vf) &&
1286                     (is_netdev == f->is_netdev)) {
1287                         f->counter--;
1288                         f->changed = true;
1289                         changed = 1;
1290                 }
1291         }
1292         if (changed) {
1293                 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1294                 vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1295                 return 0;
1296         }
1297         return -ENOENT;
1298 }
1299
1300 /**
1301  * i40e_rm_default_mac_filter - Remove the default MAC filter set by NVM
1302  * @vsi: the PF Main VSI - inappropriate for any other VSI
1303  * @macaddr: the MAC address
1304  *
1305  * Some older firmware configurations set up a default promiscuous VLAN
1306  * filter that needs to be removed.
1307  **/
1308 static int i40e_rm_default_mac_filter(struct i40e_vsi *vsi, u8 *macaddr)
1309 {
1310         struct i40e_aqc_remove_macvlan_element_data element;
1311         struct i40e_pf *pf = vsi->back;
1312         i40e_status ret;
1313
1314         /* Only appropriate for the PF main VSI */
1315         if (vsi->type != I40E_VSI_MAIN)
1316                 return -EINVAL;
1317
1318         memset(&element, 0, sizeof(element));
1319         ether_addr_copy(element.mac_addr, macaddr);
1320         element.vlan_tag = 0;
1321         element.flags = I40E_AQC_MACVLAN_DEL_PERFECT_MATCH |
1322                         I40E_AQC_MACVLAN_DEL_IGNORE_VLAN;
1323         ret = i40e_aq_remove_macvlan(&pf->hw, vsi->seid, &element, 1, NULL);
1324         if (ret)
1325                 return -ENOENT;
1326
1327         return 0;
1328 }
1329
1330 /**
1331  * i40e_add_filter - Add a mac/vlan filter to the VSI
1332  * @vsi: the VSI to be searched
1333  * @macaddr: the MAC address
1334  * @vlan: the vlan
1335  * @is_vf: make sure its a VF filter, else doesn't matter
1336  * @is_netdev: make sure its a netdev filter, else doesn't matter
1337  *
1338  * Returns ptr to the filter object or NULL when no memory available.
1339  *
1340  * NOTE: This function is expected to be called with mac_filter_list_lock
1341  * being held.
1342  **/
1343 struct i40e_mac_filter *i40e_add_filter(struct i40e_vsi *vsi,
1344                                         u8 *macaddr, s16 vlan,
1345                                         bool is_vf, bool is_netdev)
1346 {
1347         struct i40e_mac_filter *f;
1348
1349         if (!vsi || !macaddr)
1350                 return NULL;
1351
1352         f = i40e_find_filter(vsi, macaddr, vlan, is_vf, is_netdev);
1353         if (!f) {
1354                 f = kzalloc(sizeof(*f), GFP_ATOMIC);
1355                 if (!f)
1356                         goto add_filter_out;
1357
1358                 ether_addr_copy(f->macaddr, macaddr);
1359                 f->vlan = vlan;
1360                 f->changed = true;
1361
1362                 INIT_LIST_HEAD(&f->list);
1363                 list_add(&f->list, &vsi->mac_filter_list);
1364         }
1365
1366         /* increment counter and add a new flag if needed */
1367         if (is_vf) {
1368                 if (!f->is_vf) {
1369                         f->is_vf = true;
1370                         f->counter++;
1371                 }
1372         } else if (is_netdev) {
1373                 if (!f->is_netdev) {
1374                         f->is_netdev = true;
1375                         f->counter++;
1376                 }
1377         } else {
1378                 f->counter++;
1379         }
1380
1381         /* changed tells sync_filters_subtask to
1382          * push the filter down to the firmware
1383          */
1384         if (f->changed) {
1385                 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1386                 vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1387         }
1388
1389 add_filter_out:
1390         return f;
1391 }
1392
1393 /**
1394  * i40e_del_filter - Remove a mac/vlan filter from the VSI
1395  * @vsi: the VSI to be searched
1396  * @macaddr: the MAC address
1397  * @vlan: the vlan
1398  * @is_vf: make sure it's a VF filter, else doesn't matter
1399  * @is_netdev: make sure it's a netdev filter, else doesn't matter
1400  *
1401  * NOTE: This function is expected to be called with mac_filter_list_lock
1402  * being held.
1403  **/
1404 void i40e_del_filter(struct i40e_vsi *vsi,
1405                      u8 *macaddr, s16 vlan,
1406                      bool is_vf, bool is_netdev)
1407 {
1408         struct i40e_mac_filter *f;
1409
1410         if (!vsi || !macaddr)
1411                 return;
1412
1413         f = i40e_find_filter(vsi, macaddr, vlan, is_vf, is_netdev);
1414         if (!f || f->counter == 0)
1415                 return;
1416
1417         if (is_vf) {
1418                 if (f->is_vf) {
1419                         f->is_vf = false;
1420                         f->counter--;
1421                 }
1422         } else if (is_netdev) {
1423                 if (f->is_netdev) {
1424                         f->is_netdev = false;
1425                         f->counter--;
1426                 }
1427         } else {
1428                 /* make sure we don't remove a filter in use by VF or netdev */
1429                 int min_f = 0;
1430
1431                 min_f += (f->is_vf ? 1 : 0);
1432                 min_f += (f->is_netdev ? 1 : 0);
1433
1434                 if (f->counter > min_f)
1435                         f->counter--;
1436         }
1437
1438         /* counter == 0 tells sync_filters_subtask to
1439          * remove the filter from the firmware's list
1440          */
1441         if (f->counter == 0) {
1442                 f->changed = true;
1443                 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1444                 vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1445         }
1446 }
1447
1448 /**
1449  * i40e_set_mac - NDO callback to set mac address
1450  * @netdev: network interface device structure
1451  * @p: pointer to an address structure
1452  *
1453  * Returns 0 on success, negative on failure
1454  **/
1455 #ifdef I40E_FCOE
1456 int i40e_set_mac(struct net_device *netdev, void *p)
1457 #else
1458 static int i40e_set_mac(struct net_device *netdev, void *p)
1459 #endif
1460 {
1461         struct i40e_netdev_priv *np = netdev_priv(netdev);
1462         struct i40e_vsi *vsi = np->vsi;
1463         struct i40e_pf *pf = vsi->back;
1464         struct i40e_hw *hw = &pf->hw;
1465         struct sockaddr *addr = p;
1466         struct i40e_mac_filter *f;
1467
1468         if (!is_valid_ether_addr(addr->sa_data))
1469                 return -EADDRNOTAVAIL;
1470
1471         if (ether_addr_equal(netdev->dev_addr, addr->sa_data)) {
1472                 netdev_info(netdev, "already using mac address %pM\n",
1473                             addr->sa_data);
1474                 return 0;
1475         }
1476
1477         if (test_bit(__I40E_DOWN, &vsi->back->state) ||
1478             test_bit(__I40E_RESET_RECOVERY_PENDING, &vsi->back->state))
1479                 return -EADDRNOTAVAIL;
1480
1481         if (ether_addr_equal(hw->mac.addr, addr->sa_data))
1482                 netdev_info(netdev, "returning to hw mac address %pM\n",
1483                             hw->mac.addr);
1484         else
1485                 netdev_info(netdev, "set new mac address %pM\n", addr->sa_data);
1486
1487         if (vsi->type == I40E_VSI_MAIN) {
1488                 i40e_status ret;
1489
1490                 ret = i40e_aq_mac_address_write(&vsi->back->hw,
1491                                                 I40E_AQC_WRITE_TYPE_LAA_WOL,
1492                                                 addr->sa_data, NULL);
1493                 if (ret) {
1494                         netdev_info(netdev,
1495                                     "Addr change for Main VSI failed: %d\n",
1496                                     ret);
1497                         return -EADDRNOTAVAIL;
1498                 }
1499         }
1500
1501         if (ether_addr_equal(netdev->dev_addr, hw->mac.addr)) {
1502                 struct i40e_aqc_remove_macvlan_element_data element;
1503
1504                 memset(&element, 0, sizeof(element));
1505                 ether_addr_copy(element.mac_addr, netdev->dev_addr);
1506                 element.flags = I40E_AQC_MACVLAN_DEL_PERFECT_MATCH;
1507                 i40e_aq_remove_macvlan(&pf->hw, vsi->seid, &element, 1, NULL);
1508         } else {
1509                 spin_lock_bh(&vsi->mac_filter_list_lock);
1510                 i40e_del_filter(vsi, netdev->dev_addr, I40E_VLAN_ANY,
1511                                 false, false);
1512                 spin_unlock_bh(&vsi->mac_filter_list_lock);
1513         }
1514
1515         if (ether_addr_equal(addr->sa_data, hw->mac.addr)) {
1516                 struct i40e_aqc_add_macvlan_element_data element;
1517
1518                 memset(&element, 0, sizeof(element));
1519                 ether_addr_copy(element.mac_addr, hw->mac.addr);
1520                 element.flags = cpu_to_le16(I40E_AQC_MACVLAN_ADD_PERFECT_MATCH);
1521                 i40e_aq_add_macvlan(&pf->hw, vsi->seid, &element, 1, NULL);
1522         } else {
1523                 spin_lock_bh(&vsi->mac_filter_list_lock);
1524                 f = i40e_add_filter(vsi, addr->sa_data, I40E_VLAN_ANY,
1525                                     false, false);
1526                 if (f)
1527                         f->is_laa = true;
1528                 spin_unlock_bh(&vsi->mac_filter_list_lock);
1529         }
1530
1531         ether_addr_copy(netdev->dev_addr, addr->sa_data);
1532
1533         return i40e_sync_vsi_filters(vsi);
1534 }
1535
1536 /**
1537  * i40e_vsi_setup_queue_map - Setup a VSI queue map based on enabled_tc
1538  * @vsi: the VSI being setup
1539  * @ctxt: VSI context structure
1540  * @enabled_tc: Enabled TCs bitmap
1541  * @is_add: True if called before Add VSI
1542  *
1543  * Setup VSI queue mapping for enabled traffic classes.
1544  **/
1545 #ifdef I40E_FCOE
1546 void i40e_vsi_setup_queue_map(struct i40e_vsi *vsi,
1547                               struct i40e_vsi_context *ctxt,
1548                               u8 enabled_tc,
1549                               bool is_add)
1550 #else
1551 static void i40e_vsi_setup_queue_map(struct i40e_vsi *vsi,
1552                                      struct i40e_vsi_context *ctxt,
1553                                      u8 enabled_tc,
1554                                      bool is_add)
1555 #endif
1556 {
1557         struct i40e_pf *pf = vsi->back;
1558         u16 sections = 0;
1559         u8 netdev_tc = 0;
1560         u16 numtc = 0;
1561         u16 qcount;
1562         u8 offset;
1563         u16 qmap;
1564         int i;
1565         u16 num_tc_qps = 0;
1566
1567         sections = I40E_AQ_VSI_PROP_QUEUE_MAP_VALID;
1568         offset = 0;
1569
1570         if (enabled_tc && (vsi->back->flags & I40E_FLAG_DCB_ENABLED)) {
1571                 /* Find numtc from enabled TC bitmap */
1572                 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
1573                         if (enabled_tc & BIT(i)) /* TC is enabled */
1574                                 numtc++;
1575                 }
1576                 if (!numtc) {
1577                         dev_warn(&pf->pdev->dev, "DCB is enabled but no TC enabled, forcing TC0\n");
1578                         numtc = 1;
1579                 }
1580         } else {
1581                 /* At least TC0 is enabled in case of non-DCB case */
1582                 numtc = 1;
1583         }
1584
1585         vsi->tc_config.numtc = numtc;
1586         vsi->tc_config.enabled_tc = enabled_tc ? enabled_tc : 1;
1587         /* Number of queues per enabled TC */
1588         /* In MFP case we can have a much lower count of MSIx
1589          * vectors available and so we need to lower the used
1590          * q count.
1591          */
1592         if (pf->flags & I40E_FLAG_MSIX_ENABLED)
1593                 qcount = min_t(int, vsi->alloc_queue_pairs, pf->num_lan_msix);
1594         else
1595                 qcount = vsi->alloc_queue_pairs;
1596         num_tc_qps = qcount / numtc;
1597         num_tc_qps = min_t(int, num_tc_qps, i40e_pf_get_max_q_per_tc(pf));
1598
1599         /* Setup queue offset/count for all TCs for given VSI */
1600         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
1601                 /* See if the given TC is enabled for the given VSI */
1602                 if (vsi->tc_config.enabled_tc & BIT(i)) {
1603                         /* TC is enabled */
1604                         int pow, num_qps;
1605
1606                         switch (vsi->type) {
1607                         case I40E_VSI_MAIN:
1608                                 qcount = min_t(int, pf->alloc_rss_size,
1609                                                num_tc_qps);
1610                                 break;
1611 #ifdef I40E_FCOE
1612                         case I40E_VSI_FCOE:
1613                                 qcount = num_tc_qps;
1614                                 break;
1615 #endif
1616                         case I40E_VSI_FDIR:
1617                         case I40E_VSI_SRIOV:
1618                         case I40E_VSI_VMDQ2:
1619                         default:
1620                                 qcount = num_tc_qps;
1621                                 WARN_ON(i != 0);
1622                                 break;
1623                         }
1624                         vsi->tc_config.tc_info[i].qoffset = offset;
1625                         vsi->tc_config.tc_info[i].qcount = qcount;
1626
1627                         /* find the next higher power-of-2 of num queue pairs */
1628                         num_qps = qcount;
1629                         pow = 0;
1630                         while (num_qps && (BIT_ULL(pow) < qcount)) {
1631                                 pow++;
1632                                 num_qps >>= 1;
1633                         }
1634
1635                         vsi->tc_config.tc_info[i].netdev_tc = netdev_tc++;
1636                         qmap =
1637                             (offset << I40E_AQ_VSI_TC_QUE_OFFSET_SHIFT) |
1638                             (pow << I40E_AQ_VSI_TC_QUE_NUMBER_SHIFT);
1639
1640                         offset += qcount;
1641                 } else {
1642                         /* TC is not enabled so set the offset to
1643                          * default queue and allocate one queue
1644                          * for the given TC.
1645                          */
1646                         vsi->tc_config.tc_info[i].qoffset = 0;
1647                         vsi->tc_config.tc_info[i].qcount = 1;
1648                         vsi->tc_config.tc_info[i].netdev_tc = 0;
1649
1650                         qmap = 0;
1651                 }
1652                 ctxt->info.tc_mapping[i] = cpu_to_le16(qmap);
1653         }
1654
1655         /* Set actual Tx/Rx queue pairs */
1656         vsi->num_queue_pairs = offset;
1657         if ((vsi->type == I40E_VSI_MAIN) && (numtc == 1)) {
1658                 if (vsi->req_queue_pairs > 0)
1659                         vsi->num_queue_pairs = vsi->req_queue_pairs;
1660                 else if (pf->flags & I40E_FLAG_MSIX_ENABLED)
1661                         vsi->num_queue_pairs = pf->num_lan_msix;
1662         }
1663
1664         /* Scheduler section valid can only be set for ADD VSI */
1665         if (is_add) {
1666                 sections |= I40E_AQ_VSI_PROP_SCHED_VALID;
1667
1668                 ctxt->info.up_enable_bits = enabled_tc;
1669         }
1670         if (vsi->type == I40E_VSI_SRIOV) {
1671                 ctxt->info.mapping_flags |=
1672                                      cpu_to_le16(I40E_AQ_VSI_QUE_MAP_NONCONTIG);
1673                 for (i = 0; i < vsi->num_queue_pairs; i++)
1674                         ctxt->info.queue_mapping[i] =
1675                                                cpu_to_le16(vsi->base_queue + i);
1676         } else {
1677                 ctxt->info.mapping_flags |=
1678                                         cpu_to_le16(I40E_AQ_VSI_QUE_MAP_CONTIG);
1679                 ctxt->info.queue_mapping[0] = cpu_to_le16(vsi->base_queue);
1680         }
1681         ctxt->info.valid_sections |= cpu_to_le16(sections);
1682 }
1683
1684 /**
1685  * i40e_set_rx_mode - NDO callback to set the netdev filters
1686  * @netdev: network interface device structure
1687  **/
1688 #ifdef I40E_FCOE
1689 void i40e_set_rx_mode(struct net_device *netdev)
1690 #else
1691 static void i40e_set_rx_mode(struct net_device *netdev)
1692 #endif
1693 {
1694         struct i40e_netdev_priv *np = netdev_priv(netdev);
1695         struct i40e_mac_filter *f, *ftmp;
1696         struct i40e_vsi *vsi = np->vsi;
1697         struct netdev_hw_addr *uca;
1698         struct netdev_hw_addr *mca;
1699         struct netdev_hw_addr *ha;
1700
1701         spin_lock_bh(&vsi->mac_filter_list_lock);
1702
1703         /* add addr if not already in the filter list */
1704         netdev_for_each_uc_addr(uca, netdev) {
1705                 if (!i40e_find_mac(vsi, uca->addr, false, true)) {
1706                         if (i40e_is_vsi_in_vlan(vsi))
1707                                 i40e_put_mac_in_vlan(vsi, uca->addr,
1708                                                      false, true);
1709                         else
1710                                 i40e_add_filter(vsi, uca->addr, I40E_VLAN_ANY,
1711                                                 false, true);
1712                 }
1713         }
1714
1715         netdev_for_each_mc_addr(mca, netdev) {
1716                 if (!i40e_find_mac(vsi, mca->addr, false, true)) {
1717                         if (i40e_is_vsi_in_vlan(vsi))
1718                                 i40e_put_mac_in_vlan(vsi, mca->addr,
1719                                                      false, true);
1720                         else
1721                                 i40e_add_filter(vsi, mca->addr, I40E_VLAN_ANY,
1722                                                 false, true);
1723                 }
1724         }
1725
1726         /* remove filter if not in netdev list */
1727         list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
1728
1729                 if (!f->is_netdev)
1730                         continue;
1731
1732                 netdev_for_each_mc_addr(mca, netdev)
1733                         if (ether_addr_equal(mca->addr, f->macaddr))
1734                                 goto bottom_of_search_loop;
1735
1736                 netdev_for_each_uc_addr(uca, netdev)
1737                         if (ether_addr_equal(uca->addr, f->macaddr))
1738                                 goto bottom_of_search_loop;
1739
1740                 for_each_dev_addr(netdev, ha)
1741                         if (ether_addr_equal(ha->addr, f->macaddr))
1742                                 goto bottom_of_search_loop;
1743
1744                 /* f->macaddr wasn't found in uc, mc, or ha list so delete it */
1745                 i40e_del_filter(vsi, f->macaddr, I40E_VLAN_ANY, false, true);
1746
1747 bottom_of_search_loop:
1748                 continue;
1749         }
1750         spin_unlock_bh(&vsi->mac_filter_list_lock);
1751
1752         /* check for other flag changes */
1753         if (vsi->current_netdev_flags != vsi->netdev->flags) {
1754                 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1755                 vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1756         }
1757 }
1758
1759 /**
1760  * i40e_mac_filter_entry_clone - Clones a MAC filter entry
1761  * @src: source MAC filter entry to be clones
1762  *
1763  * Returns the pointer to newly cloned MAC filter entry or NULL
1764  * in case of error
1765  **/
1766 static struct i40e_mac_filter *i40e_mac_filter_entry_clone(
1767                                         struct i40e_mac_filter *src)
1768 {
1769         struct i40e_mac_filter *f;
1770
1771         f = kzalloc(sizeof(*f), GFP_ATOMIC);
1772         if (!f)
1773                 return NULL;
1774         *f = *src;
1775
1776         INIT_LIST_HEAD(&f->list);
1777
1778         return f;
1779 }
1780
1781 /**
1782  * i40e_undo_del_filter_entries - Undo the changes made to MAC filter entries
1783  * @vsi: pointer to vsi struct
1784  * @from: Pointer to list which contains MAC filter entries - changes to
1785  *        those entries needs to be undone.
1786  *
1787  * MAC filter entries from list were slated to be removed from device.
1788  **/
1789 static void i40e_undo_del_filter_entries(struct i40e_vsi *vsi,
1790                                          struct list_head *from)
1791 {
1792         struct i40e_mac_filter *f, *ftmp;
1793
1794         list_for_each_entry_safe(f, ftmp, from, list) {
1795                 f->changed = true;
1796                 /* Move the element back into MAC filter list*/
1797                 list_move_tail(&f->list, &vsi->mac_filter_list);
1798         }
1799 }
1800
1801 /**
1802  * i40e_undo_add_filter_entries - Undo the changes made to MAC filter entries
1803  * @vsi: pointer to vsi struct
1804  *
1805  * MAC filter entries from list were slated to be added from device.
1806  **/
1807 static void i40e_undo_add_filter_entries(struct i40e_vsi *vsi)
1808 {
1809         struct i40e_mac_filter *f, *ftmp;
1810
1811         list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
1812                 if (!f->changed && f->counter)
1813                         f->changed = true;
1814         }
1815 }
1816
1817 /**
1818  * i40e_cleanup_add_list - Deletes the element from add list and release
1819  *                      memory
1820  * @add_list: Pointer to list which contains MAC filter entries
1821  **/
1822 static void i40e_cleanup_add_list(struct list_head *add_list)
1823 {
1824         struct i40e_mac_filter *f, *ftmp;
1825
1826         list_for_each_entry_safe(f, ftmp, add_list, list) {
1827                 list_del(&f->list);
1828                 kfree(f);
1829         }
1830 }
1831
1832 /**
1833  * i40e_sync_vsi_filters - Update the VSI filter list to the HW
1834  * @vsi: ptr to the VSI
1835  *
1836  * Push any outstanding VSI filter changes through the AdminQ.
1837  *
1838  * Returns 0 or error value
1839  **/
1840 int i40e_sync_vsi_filters(struct i40e_vsi *vsi)
1841 {
1842         struct list_head tmp_del_list, tmp_add_list;
1843         struct i40e_mac_filter *f, *ftmp, *fclone;
1844         bool promisc_forced_on = false;
1845         bool add_happened = false;
1846         int filter_list_len = 0;
1847         u32 changed_flags = 0;
1848         i40e_status aq_ret = 0;
1849         bool err_cond = false;
1850         int retval = 0;
1851         struct i40e_pf *pf;
1852         int num_add = 0;
1853         int num_del = 0;
1854         int aq_err = 0;
1855         u16 cmd_flags;
1856
1857         /* empty array typed pointers, kcalloc later */
1858         struct i40e_aqc_add_macvlan_element_data *add_list;
1859         struct i40e_aqc_remove_macvlan_element_data *del_list;
1860
1861         while (test_and_set_bit(__I40E_CONFIG_BUSY, &vsi->state))
1862                 usleep_range(1000, 2000);
1863         pf = vsi->back;
1864
1865         if (vsi->netdev) {
1866                 changed_flags = vsi->current_netdev_flags ^ vsi->netdev->flags;
1867                 vsi->current_netdev_flags = vsi->netdev->flags;
1868         }
1869
1870         INIT_LIST_HEAD(&tmp_del_list);
1871         INIT_LIST_HEAD(&tmp_add_list);
1872
1873         if (vsi->flags & I40E_VSI_FLAG_FILTER_CHANGED) {
1874                 vsi->flags &= ~I40E_VSI_FLAG_FILTER_CHANGED;
1875
1876                 spin_lock_bh(&vsi->mac_filter_list_lock);
1877                 list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
1878                         if (!f->changed)
1879                                 continue;
1880
1881                         if (f->counter != 0)
1882                                 continue;
1883                         f->changed = false;
1884
1885                         /* Move the element into temporary del_list */
1886                         list_move_tail(&f->list, &tmp_del_list);
1887                 }
1888
1889                 list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
1890                         if (!f->changed)
1891                                 continue;
1892
1893                         if (f->counter == 0)
1894                                 continue;
1895                         f->changed = false;
1896
1897                         /* Clone MAC filter entry and add into temporary list */
1898                         fclone = i40e_mac_filter_entry_clone(f);
1899                         if (!fclone) {
1900                                 err_cond = true;
1901                                 break;
1902                         }
1903                         list_add_tail(&fclone->list, &tmp_add_list);
1904                 }
1905
1906                 /* if failed to clone MAC filter entry - undo */
1907                 if (err_cond) {
1908                         i40e_undo_del_filter_entries(vsi, &tmp_del_list);
1909                         i40e_undo_add_filter_entries(vsi);
1910                 }
1911                 spin_unlock_bh(&vsi->mac_filter_list_lock);
1912
1913                 if (err_cond) {
1914                         i40e_cleanup_add_list(&tmp_add_list);
1915                         retval = -ENOMEM;
1916                         goto out;
1917                 }
1918         }
1919
1920         /* Now process 'del_list' outside the lock */
1921         if (!list_empty(&tmp_del_list)) {
1922                 int del_list_size;
1923
1924                 filter_list_len = pf->hw.aq.asq_buf_size /
1925                             sizeof(struct i40e_aqc_remove_macvlan_element_data);
1926                 del_list_size = filter_list_len *
1927                             sizeof(struct i40e_aqc_remove_macvlan_element_data);
1928                 del_list = kzalloc(del_list_size, GFP_KERNEL);
1929                 if (!del_list) {
1930                         i40e_cleanup_add_list(&tmp_add_list);
1931
1932                         /* Undo VSI's MAC filter entry element updates */
1933                         spin_lock_bh(&vsi->mac_filter_list_lock);
1934                         i40e_undo_del_filter_entries(vsi, &tmp_del_list);
1935                         i40e_undo_add_filter_entries(vsi);
1936                         spin_unlock_bh(&vsi->mac_filter_list_lock);
1937                         retval = -ENOMEM;
1938                         goto out;
1939                 }
1940
1941                 list_for_each_entry_safe(f, ftmp, &tmp_del_list, list) {
1942                         cmd_flags = 0;
1943
1944                         /* add to delete list */
1945                         ether_addr_copy(del_list[num_del].mac_addr, f->macaddr);
1946                         del_list[num_del].vlan_tag =
1947                                 cpu_to_le16((u16)(f->vlan ==
1948                                             I40E_VLAN_ANY ? 0 : f->vlan));
1949
1950                         cmd_flags |= I40E_AQC_MACVLAN_DEL_PERFECT_MATCH;
1951                         del_list[num_del].flags = cmd_flags;
1952                         num_del++;
1953
1954                         /* flush a full buffer */
1955                         if (num_del == filter_list_len) {
1956                                 aq_ret = i40e_aq_remove_macvlan(&pf->hw,
1957                                                                 vsi->seid,
1958                                                                 del_list,
1959                                                                 num_del,
1960                                                                 NULL);
1961                                 aq_err = pf->hw.aq.asq_last_status;
1962                                 num_del = 0;
1963                                 memset(del_list, 0, del_list_size);
1964
1965                                 if (aq_ret && aq_err != I40E_AQ_RC_ENOENT) {
1966                                         retval = -EIO;
1967                                         dev_err(&pf->pdev->dev,
1968                                                 "ignoring delete macvlan error, err %s, aq_err %s while flushing a full buffer\n",
1969                                                 i40e_stat_str(&pf->hw, aq_ret),
1970                                                 i40e_aq_str(&pf->hw, aq_err));
1971                                 }
1972                         }
1973                         /* Release memory for MAC filter entries which were
1974                          * synced up with HW.
1975                          */
1976                         list_del(&f->list);
1977                         kfree(f);
1978                 }
1979
1980                 if (num_del) {
1981                         aq_ret = i40e_aq_remove_macvlan(&pf->hw, vsi->seid,
1982                                                         del_list, num_del,
1983                                                         NULL);
1984                         aq_err = pf->hw.aq.asq_last_status;
1985                         num_del = 0;
1986
1987                         if (aq_ret && aq_err != I40E_AQ_RC_ENOENT)
1988                                 dev_info(&pf->pdev->dev,
1989                                          "ignoring delete macvlan error, err %s aq_err %s\n",
1990                                          i40e_stat_str(&pf->hw, aq_ret),
1991                                          i40e_aq_str(&pf->hw, aq_err));
1992                 }
1993
1994                 kfree(del_list);
1995                 del_list = NULL;
1996         }
1997
1998         if (!list_empty(&tmp_add_list)) {
1999                 int add_list_size;
2000
2001                 /* do all the adds now */
2002                 filter_list_len = pf->hw.aq.asq_buf_size /
2003                                sizeof(struct i40e_aqc_add_macvlan_element_data),
2004                 add_list_size = filter_list_len *
2005                                sizeof(struct i40e_aqc_add_macvlan_element_data);
2006                 add_list = kzalloc(add_list_size, GFP_KERNEL);
2007                 if (!add_list) {
2008                         /* Purge element from temporary lists */
2009                         i40e_cleanup_add_list(&tmp_add_list);
2010
2011                         /* Undo add filter entries from VSI MAC filter list */
2012                         spin_lock_bh(&vsi->mac_filter_list_lock);
2013                         i40e_undo_add_filter_entries(vsi);
2014                         spin_unlock_bh(&vsi->mac_filter_list_lock);
2015                         retval = -ENOMEM;
2016                         goto out;
2017                 }
2018
2019                 list_for_each_entry_safe(f, ftmp, &tmp_add_list, list) {
2020
2021                         add_happened = true;
2022                         cmd_flags = 0;
2023
2024                         /* add to add array */
2025                         ether_addr_copy(add_list[num_add].mac_addr, f->macaddr);
2026                         add_list[num_add].vlan_tag =
2027                                 cpu_to_le16(
2028                                  (u16)(f->vlan == I40E_VLAN_ANY ? 0 : f->vlan));
2029                         add_list[num_add].queue_number = 0;
2030
2031                         cmd_flags |= I40E_AQC_MACVLAN_ADD_PERFECT_MATCH;
2032                         add_list[num_add].flags = cpu_to_le16(cmd_flags);
2033                         num_add++;
2034
2035                         /* flush a full buffer */
2036                         if (num_add == filter_list_len) {
2037                                 aq_ret = i40e_aq_add_macvlan(&pf->hw, vsi->seid,
2038                                                              add_list, num_add,
2039                                                              NULL);
2040                                 aq_err = pf->hw.aq.asq_last_status;
2041                                 num_add = 0;
2042
2043                                 if (aq_ret)
2044                                         break;
2045                                 memset(add_list, 0, add_list_size);
2046                         }
2047                         /* Entries from tmp_add_list were cloned from MAC
2048                          * filter list, hence clean those cloned entries
2049                          */
2050                         list_del(&f->list);
2051                         kfree(f);
2052                 }
2053
2054                 if (num_add) {
2055                         aq_ret = i40e_aq_add_macvlan(&pf->hw, vsi->seid,
2056                                                      add_list, num_add, NULL);
2057                         aq_err = pf->hw.aq.asq_last_status;
2058                         num_add = 0;
2059                 }
2060                 kfree(add_list);
2061                 add_list = NULL;
2062
2063                 if (add_happened && aq_ret && aq_err != I40E_AQ_RC_EINVAL) {
2064                         retval = i40e_aq_rc_to_posix(aq_ret, aq_err);
2065                         dev_info(&pf->pdev->dev,
2066                                  "add filter failed, err %s aq_err %s\n",
2067                                  i40e_stat_str(&pf->hw, aq_ret),
2068                                  i40e_aq_str(&pf->hw, aq_err));
2069                         if ((pf->hw.aq.asq_last_status == I40E_AQ_RC_ENOSPC) &&
2070                             !test_bit(__I40E_FILTER_OVERFLOW_PROMISC,
2071                                       &vsi->state)) {
2072                                 promisc_forced_on = true;
2073                                 set_bit(__I40E_FILTER_OVERFLOW_PROMISC,
2074                                         &vsi->state);
2075                                 dev_info(&pf->pdev->dev, "promiscuous mode forced on\n");
2076                         }
2077                 }
2078         }
2079
2080         /* check for changes in promiscuous modes */
2081         if (changed_flags & IFF_ALLMULTI) {
2082                 bool cur_multipromisc;
2083
2084                 cur_multipromisc = !!(vsi->current_netdev_flags & IFF_ALLMULTI);
2085                 aq_ret = i40e_aq_set_vsi_multicast_promiscuous(&vsi->back->hw,
2086                                                                vsi->seid,
2087                                                                cur_multipromisc,
2088                                                                NULL);
2089                 if (aq_ret) {
2090                         retval = i40e_aq_rc_to_posix(aq_ret,
2091                                                      pf->hw.aq.asq_last_status);
2092                         dev_info(&pf->pdev->dev,
2093                                  "set multi promisc failed, err %s aq_err %s\n",
2094                                  i40e_stat_str(&pf->hw, aq_ret),
2095                                  i40e_aq_str(&pf->hw,
2096                                              pf->hw.aq.asq_last_status));
2097                 }
2098         }
2099         if ((changed_flags & IFF_PROMISC) || promisc_forced_on) {
2100                 bool cur_promisc;
2101
2102                 cur_promisc = (!!(vsi->current_netdev_flags & IFF_PROMISC) ||
2103                                test_bit(__I40E_FILTER_OVERFLOW_PROMISC,
2104                                         &vsi->state));
2105                 if (vsi->type == I40E_VSI_MAIN && pf->lan_veb != I40E_NO_VEB) {
2106                         /* set defport ON for Main VSI instead of true promisc
2107                          * this way we will get all unicast/multicast and VLAN
2108                          * promisc behavior but will not get VF or VMDq traffic
2109                          * replicated on the Main VSI.
2110                          */
2111                         if (pf->cur_promisc != cur_promisc) {
2112                                 pf->cur_promisc = cur_promisc;
2113                                 set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
2114                         }
2115                 } else {
2116                         aq_ret = i40e_aq_set_vsi_unicast_promiscuous(
2117                                                           &vsi->back->hw,
2118                                                           vsi->seid,
2119                                                           cur_promisc, NULL);
2120                         if (aq_ret) {
2121                                 retval =
2122                                 i40e_aq_rc_to_posix(aq_ret,
2123                                                     pf->hw.aq.asq_last_status);
2124                                 dev_info(&pf->pdev->dev,
2125                                          "set unicast promisc failed, err %d, aq_err %d\n",
2126                                          aq_ret, pf->hw.aq.asq_last_status);
2127                         }
2128                         aq_ret = i40e_aq_set_vsi_multicast_promiscuous(
2129                                                           &vsi->back->hw,
2130                                                           vsi->seid,
2131                                                           cur_promisc, NULL);
2132                         if (aq_ret) {
2133                                 retval =
2134                                 i40e_aq_rc_to_posix(aq_ret,
2135                                                     pf->hw.aq.asq_last_status);
2136                                 dev_info(&pf->pdev->dev,
2137                                          "set multicast promisc failed, err %d, aq_err %d\n",
2138                                          aq_ret, pf->hw.aq.asq_last_status);
2139                         }
2140                 }
2141                 aq_ret = i40e_aq_set_vsi_broadcast(&vsi->back->hw,
2142                                                    vsi->seid,
2143                                                    cur_promisc, NULL);
2144                 if (aq_ret) {
2145                         retval = i40e_aq_rc_to_posix(aq_ret,
2146                                                      pf->hw.aq.asq_last_status);
2147                         dev_info(&pf->pdev->dev,
2148                                  "set brdcast promisc failed, err %s, aq_err %s\n",
2149                                  i40e_stat_str(&pf->hw, aq_ret),
2150                                  i40e_aq_str(&pf->hw,
2151                                              pf->hw.aq.asq_last_status));
2152                 }
2153         }
2154 out:
2155         clear_bit(__I40E_CONFIG_BUSY, &vsi->state);
2156         return retval;
2157 }
2158
2159 /**
2160  * i40e_sync_filters_subtask - Sync the VSI filter list with HW
2161  * @pf: board private structure
2162  **/
2163 static void i40e_sync_filters_subtask(struct i40e_pf *pf)
2164 {
2165         int v;
2166
2167         if (!pf || !(pf->flags & I40E_FLAG_FILTER_SYNC))
2168                 return;
2169         pf->flags &= ~I40E_FLAG_FILTER_SYNC;
2170
2171         for (v = 0; v < pf->num_alloc_vsi; v++) {
2172                 if (pf->vsi[v] &&
2173                     (pf->vsi[v]->flags & I40E_VSI_FLAG_FILTER_CHANGED)) {
2174                         int ret = i40e_sync_vsi_filters(pf->vsi[v]);
2175
2176                         if (ret) {
2177                                 /* come back and try again later */
2178                                 pf->flags |= I40E_FLAG_FILTER_SYNC;
2179                                 break;
2180                         }
2181                 }
2182         }
2183 }
2184
2185 /**
2186  * i40e_change_mtu - NDO callback to change the Maximum Transfer Unit
2187  * @netdev: network interface device structure
2188  * @new_mtu: new value for maximum frame size
2189  *
2190  * Returns 0 on success, negative on failure
2191  **/
2192 static int i40e_change_mtu(struct net_device *netdev, int new_mtu)
2193 {
2194         struct i40e_netdev_priv *np = netdev_priv(netdev);
2195         int max_frame = new_mtu + ETH_HLEN + ETH_FCS_LEN + VLAN_HLEN;
2196         struct i40e_vsi *vsi = np->vsi;
2197
2198         /* MTU < 68 is an error and causes problems on some kernels */
2199         if ((new_mtu < 68) || (max_frame > I40E_MAX_RXBUFFER))
2200                 return -EINVAL;
2201
2202         netdev_info(netdev, "changing MTU from %d to %d\n",
2203                     netdev->mtu, new_mtu);
2204         netdev->mtu = new_mtu;
2205         if (netif_running(netdev))
2206                 i40e_vsi_reinit_locked(vsi);
2207
2208         return 0;
2209 }
2210
2211 /**
2212  * i40e_ioctl - Access the hwtstamp interface
2213  * @netdev: network interface device structure
2214  * @ifr: interface request data
2215  * @cmd: ioctl command
2216  **/
2217 int i40e_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
2218 {
2219         struct i40e_netdev_priv *np = netdev_priv(netdev);
2220         struct i40e_pf *pf = np->vsi->back;
2221
2222         switch (cmd) {
2223         case SIOCGHWTSTAMP:
2224                 return i40e_ptp_get_ts_config(pf, ifr);
2225         case SIOCSHWTSTAMP:
2226                 return i40e_ptp_set_ts_config(pf, ifr);
2227         default:
2228                 return -EOPNOTSUPP;
2229         }
2230 }
2231
2232 /**
2233  * i40e_vlan_stripping_enable - Turn on vlan stripping for the VSI
2234  * @vsi: the vsi being adjusted
2235  **/
2236 void i40e_vlan_stripping_enable(struct i40e_vsi *vsi)
2237 {
2238         struct i40e_vsi_context ctxt;
2239         i40e_status ret;
2240
2241         if ((vsi->info.valid_sections &
2242              cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID)) &&
2243             ((vsi->info.port_vlan_flags & I40E_AQ_VSI_PVLAN_MODE_MASK) == 0))
2244                 return;  /* already enabled */
2245
2246         vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2247         vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_ALL |
2248                                     I40E_AQ_VSI_PVLAN_EMOD_STR_BOTH;
2249
2250         ctxt.seid = vsi->seid;
2251         ctxt.info = vsi->info;
2252         ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2253         if (ret) {
2254                 dev_info(&vsi->back->pdev->dev,
2255                          "update vlan stripping failed, err %s aq_err %s\n",
2256                          i40e_stat_str(&vsi->back->hw, ret),
2257                          i40e_aq_str(&vsi->back->hw,
2258                                      vsi->back->hw.aq.asq_last_status));
2259         }
2260 }
2261
2262 /**
2263  * i40e_vlan_stripping_disable - Turn off vlan stripping for the VSI
2264  * @vsi: the vsi being adjusted
2265  **/
2266 void i40e_vlan_stripping_disable(struct i40e_vsi *vsi)
2267 {
2268         struct i40e_vsi_context ctxt;
2269         i40e_status ret;
2270
2271         if ((vsi->info.valid_sections &
2272              cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID)) &&
2273             ((vsi->info.port_vlan_flags & I40E_AQ_VSI_PVLAN_EMOD_MASK) ==
2274              I40E_AQ_VSI_PVLAN_EMOD_MASK))
2275                 return;  /* already disabled */
2276
2277         vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2278         vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_ALL |
2279                                     I40E_AQ_VSI_PVLAN_EMOD_NOTHING;
2280
2281         ctxt.seid = vsi->seid;
2282         ctxt.info = vsi->info;
2283         ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2284         if (ret) {
2285                 dev_info(&vsi->back->pdev->dev,
2286                          "update vlan stripping failed, err %s aq_err %s\n",
2287                          i40e_stat_str(&vsi->back->hw, ret),
2288                          i40e_aq_str(&vsi->back->hw,
2289                                      vsi->back->hw.aq.asq_last_status));
2290         }
2291 }
2292
2293 /**
2294  * i40e_vlan_rx_register - Setup or shutdown vlan offload
2295  * @netdev: network interface to be adjusted
2296  * @features: netdev features to test if VLAN offload is enabled or not
2297  **/
2298 static void i40e_vlan_rx_register(struct net_device *netdev, u32 features)
2299 {
2300         struct i40e_netdev_priv *np = netdev_priv(netdev);
2301         struct i40e_vsi *vsi = np->vsi;
2302
2303         if (features & NETIF_F_HW_VLAN_CTAG_RX)
2304                 i40e_vlan_stripping_enable(vsi);
2305         else
2306                 i40e_vlan_stripping_disable(vsi);
2307 }
2308
2309 /**
2310  * i40e_vsi_add_vlan - Add vsi membership for given vlan
2311  * @vsi: the vsi being configured
2312  * @vid: vlan id to be added (0 = untagged only , -1 = any)
2313  **/
2314 int i40e_vsi_add_vlan(struct i40e_vsi *vsi, s16 vid)
2315 {
2316         struct i40e_mac_filter *f, *add_f;
2317         bool is_netdev, is_vf;
2318
2319         is_vf = (vsi->type == I40E_VSI_SRIOV);
2320         is_netdev = !!(vsi->netdev);
2321
2322         /* Locked once because all functions invoked below iterates list*/
2323         spin_lock_bh(&vsi->mac_filter_list_lock);
2324
2325         if (is_netdev) {
2326                 add_f = i40e_add_filter(vsi, vsi->netdev->dev_addr, vid,
2327                                         is_vf, is_netdev);
2328                 if (!add_f) {
2329                         dev_info(&vsi->back->pdev->dev,
2330                                  "Could not add vlan filter %d for %pM\n",
2331                                  vid, vsi->netdev->dev_addr);
2332                         spin_unlock_bh(&vsi->mac_filter_list_lock);
2333                         return -ENOMEM;
2334                 }
2335         }
2336
2337         list_for_each_entry(f, &vsi->mac_filter_list, list) {
2338                 add_f = i40e_add_filter(vsi, f->macaddr, vid, is_vf, is_netdev);
2339                 if (!add_f) {
2340                         dev_info(&vsi->back->pdev->dev,
2341                                  "Could not add vlan filter %d for %pM\n",
2342                                  vid, f->macaddr);
2343                         spin_unlock_bh(&vsi->mac_filter_list_lock);
2344                         return -ENOMEM;
2345                 }
2346         }
2347
2348         /* Now if we add a vlan tag, make sure to check if it is the first
2349          * tag (i.e. a "tag" -1 does exist) and if so replace the -1 "tag"
2350          * with 0, so we now accept untagged and specified tagged traffic
2351          * (and not any taged and untagged)
2352          */
2353         if (vid > 0) {
2354                 if (is_netdev && i40e_find_filter(vsi, vsi->netdev->dev_addr,
2355                                                   I40E_VLAN_ANY,
2356                                                   is_vf, is_netdev)) {
2357                         i40e_del_filter(vsi, vsi->netdev->dev_addr,
2358                                         I40E_VLAN_ANY, is_vf, is_netdev);
2359                         add_f = i40e_add_filter(vsi, vsi->netdev->dev_addr, 0,
2360                                                 is_vf, is_netdev);
2361                         if (!add_f) {
2362                                 dev_info(&vsi->back->pdev->dev,
2363                                          "Could not add filter 0 for %pM\n",
2364                                          vsi->netdev->dev_addr);
2365                                 spin_unlock_bh(&vsi->mac_filter_list_lock);
2366                                 return -ENOMEM;
2367                         }
2368                 }
2369         }
2370
2371         /* Do not assume that I40E_VLAN_ANY should be reset to VLAN 0 */
2372         if (vid > 0 && !vsi->info.pvid) {
2373                 list_for_each_entry(f, &vsi->mac_filter_list, list) {
2374                         if (!i40e_find_filter(vsi, f->macaddr, I40E_VLAN_ANY,
2375                                               is_vf, is_netdev))
2376                                 continue;
2377                         i40e_del_filter(vsi, f->macaddr, I40E_VLAN_ANY,
2378                                         is_vf, is_netdev);
2379                         add_f = i40e_add_filter(vsi, f->macaddr,
2380                                                 0, is_vf, is_netdev);
2381                         if (!add_f) {
2382                                 dev_info(&vsi->back->pdev->dev,
2383                                          "Could not add filter 0 for %pM\n",
2384                                         f->macaddr);
2385                                 spin_unlock_bh(&vsi->mac_filter_list_lock);
2386                                 return -ENOMEM;
2387                         }
2388                 }
2389         }
2390
2391         spin_unlock_bh(&vsi->mac_filter_list_lock);
2392
2393         /* schedule our worker thread which will take care of
2394          * applying the new filter changes
2395          */
2396         i40e_service_event_schedule(vsi->back);
2397         return 0;
2398 }
2399
2400 /**
2401  * i40e_vsi_kill_vlan - Remove vsi membership for given vlan
2402  * @vsi: the vsi being configured
2403  * @vid: vlan id to be removed (0 = untagged only , -1 = any)
2404  *
2405  * Return: 0 on success or negative otherwise
2406  **/
2407 int i40e_vsi_kill_vlan(struct i40e_vsi *vsi, s16 vid)
2408 {
2409         struct net_device *netdev = vsi->netdev;
2410         struct i40e_mac_filter *f, *add_f;
2411         bool is_vf, is_netdev;
2412         int filter_count = 0;
2413
2414         is_vf = (vsi->type == I40E_VSI_SRIOV);
2415         is_netdev = !!(netdev);
2416
2417         /* Locked once because all functions invoked below iterates list */
2418         spin_lock_bh(&vsi->mac_filter_list_lock);
2419
2420         if (is_netdev)
2421                 i40e_del_filter(vsi, netdev->dev_addr, vid, is_vf, is_netdev);
2422
2423         list_for_each_entry(f, &vsi->mac_filter_list, list)
2424                 i40e_del_filter(vsi, f->macaddr, vid, is_vf, is_netdev);
2425
2426         /* go through all the filters for this VSI and if there is only
2427          * vid == 0 it means there are no other filters, so vid 0 must
2428          * be replaced with -1. This signifies that we should from now
2429          * on accept any traffic (with any tag present, or untagged)
2430          */
2431         list_for_each_entry(f, &vsi->mac_filter_list, list) {
2432                 if (is_netdev) {
2433                         if (f->vlan &&
2434                             ether_addr_equal(netdev->dev_addr, f->macaddr))
2435                                 filter_count++;
2436                 }
2437
2438                 if (f->vlan)
2439                         filter_count++;
2440         }
2441
2442         if (!filter_count && is_netdev) {
2443                 i40e_del_filter(vsi, netdev->dev_addr, 0, is_vf, is_netdev);
2444                 f = i40e_add_filter(vsi, netdev->dev_addr, I40E_VLAN_ANY,
2445                                     is_vf, is_netdev);
2446                 if (!f) {
2447                         dev_info(&vsi->back->pdev->dev,
2448                                  "Could not add filter %d for %pM\n",
2449                                  I40E_VLAN_ANY, netdev->dev_addr);
2450                         spin_unlock_bh(&vsi->mac_filter_list_lock);
2451                         return -ENOMEM;
2452                 }
2453         }
2454
2455         if (!filter_count) {
2456                 list_for_each_entry(f, &vsi->mac_filter_list, list) {
2457                         i40e_del_filter(vsi, f->macaddr, 0, is_vf, is_netdev);
2458                         add_f = i40e_add_filter(vsi, f->macaddr, I40E_VLAN_ANY,
2459                                                 is_vf, is_netdev);
2460                         if (!add_f) {
2461                                 dev_info(&vsi->back->pdev->dev,
2462                                          "Could not add filter %d for %pM\n",
2463                                          I40E_VLAN_ANY, f->macaddr);
2464                                 spin_unlock_bh(&vsi->mac_filter_list_lock);
2465                                 return -ENOMEM;
2466                         }
2467                 }
2468         }
2469
2470         spin_unlock_bh(&vsi->mac_filter_list_lock);
2471
2472         /* schedule our worker thread which will take care of
2473          * applying the new filter changes
2474          */
2475         i40e_service_event_schedule(vsi->back);
2476         return 0;
2477 }
2478
2479 /**
2480  * i40e_vlan_rx_add_vid - Add a vlan id filter to HW offload
2481  * @netdev: network interface to be adjusted
2482  * @vid: vlan id to be added
2483  *
2484  * net_device_ops implementation for adding vlan ids
2485  **/
2486 #ifdef I40E_FCOE
2487 int i40e_vlan_rx_add_vid(struct net_device *netdev,
2488                          __always_unused __be16 proto, u16 vid)
2489 #else
2490 static int i40e_vlan_rx_add_vid(struct net_device *netdev,
2491                                 __always_unused __be16 proto, u16 vid)
2492 #endif
2493 {
2494         struct i40e_netdev_priv *np = netdev_priv(netdev);
2495         struct i40e_vsi *vsi = np->vsi;
2496         int ret = 0;
2497
2498         if (vid > 4095)
2499                 return -EINVAL;
2500
2501         netdev_info(netdev, "adding %pM vid=%d\n", netdev->dev_addr, vid);
2502
2503         /* If the network stack called us with vid = 0 then
2504          * it is asking to receive priority tagged packets with
2505          * vlan id 0.  Our HW receives them by default when configured
2506          * to receive untagged packets so there is no need to add an
2507          * extra filter for vlan 0 tagged packets.
2508          */
2509         if (vid)
2510                 ret = i40e_vsi_add_vlan(vsi, vid);
2511
2512         if (!ret && (vid < VLAN_N_VID))
2513                 set_bit(vid, vsi->active_vlans);
2514
2515         return ret;
2516 }
2517
2518 /**
2519  * i40e_vlan_rx_kill_vid - Remove a vlan id filter from HW offload
2520  * @netdev: network interface to be adjusted
2521  * @vid: vlan id to be removed
2522  *
2523  * net_device_ops implementation for removing vlan ids
2524  **/
2525 #ifdef I40E_FCOE
2526 int i40e_vlan_rx_kill_vid(struct net_device *netdev,
2527                           __always_unused __be16 proto, u16 vid)
2528 #else
2529 static int i40e_vlan_rx_kill_vid(struct net_device *netdev,
2530                                  __always_unused __be16 proto, u16 vid)
2531 #endif
2532 {
2533         struct i40e_netdev_priv *np = netdev_priv(netdev);
2534         struct i40e_vsi *vsi = np->vsi;
2535
2536         netdev_info(netdev, "removing %pM vid=%d\n", netdev->dev_addr, vid);
2537
2538         /* return code is ignored as there is nothing a user
2539          * can do about failure to remove and a log message was
2540          * already printed from the other function
2541          */
2542         i40e_vsi_kill_vlan(vsi, vid);
2543
2544         clear_bit(vid, vsi->active_vlans);
2545
2546         return 0;
2547 }
2548
2549 /**
2550  * i40e_restore_vlan - Reinstate vlans when vsi/netdev comes back up
2551  * @vsi: the vsi being brought back up
2552  **/
2553 static void i40e_restore_vlan(struct i40e_vsi *vsi)
2554 {
2555         u16 vid;
2556
2557         if (!vsi->netdev)
2558                 return;
2559
2560         i40e_vlan_rx_register(vsi->netdev, vsi->netdev->features);
2561
2562         for_each_set_bit(vid, vsi->active_vlans, VLAN_N_VID)
2563                 i40e_vlan_rx_add_vid(vsi->netdev, htons(ETH_P_8021Q),
2564                                      vid);
2565 }
2566
2567 /**
2568  * i40e_vsi_add_pvid - Add pvid for the VSI
2569  * @vsi: the vsi being adjusted
2570  * @vid: the vlan id to set as a PVID
2571  **/
2572 int i40e_vsi_add_pvid(struct i40e_vsi *vsi, u16 vid)
2573 {
2574         struct i40e_vsi_context ctxt;
2575         i40e_status ret;
2576
2577         vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2578         vsi->info.pvid = cpu_to_le16(vid);
2579         vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_TAGGED |
2580                                     I40E_AQ_VSI_PVLAN_INSERT_PVID |
2581                                     I40E_AQ_VSI_PVLAN_EMOD_STR;
2582
2583         ctxt.seid = vsi->seid;
2584         ctxt.info = vsi->info;
2585         ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2586         if (ret) {
2587                 dev_info(&vsi->back->pdev->dev,
2588                          "add pvid failed, err %s aq_err %s\n",
2589                          i40e_stat_str(&vsi->back->hw, ret),
2590                          i40e_aq_str(&vsi->back->hw,
2591                                      vsi->back->hw.aq.asq_last_status));
2592                 return -ENOENT;
2593         }
2594
2595         return 0;
2596 }
2597
2598 /**
2599  * i40e_vsi_remove_pvid - Remove the pvid from the VSI
2600  * @vsi: the vsi being adjusted
2601  *
2602  * Just use the vlan_rx_register() service to put it back to normal
2603  **/
2604 void i40e_vsi_remove_pvid(struct i40e_vsi *vsi)
2605 {
2606         i40e_vlan_stripping_disable(vsi);
2607
2608         vsi->info.pvid = 0;
2609 }
2610
2611 /**
2612  * i40e_vsi_setup_tx_resources - Allocate VSI Tx queue resources
2613  * @vsi: ptr to the VSI
2614  *
2615  * If this function returns with an error, then it's possible one or
2616  * more of the rings is populated (while the rest are not).  It is the
2617  * callers duty to clean those orphaned rings.
2618  *
2619  * Return 0 on success, negative on failure
2620  **/
2621 static int i40e_vsi_setup_tx_resources(struct i40e_vsi *vsi)
2622 {
2623         int i, err = 0;
2624
2625         for (i = 0; i < vsi->num_queue_pairs && !err; i++)
2626                 err = i40e_setup_tx_descriptors(vsi->tx_rings[i]);
2627
2628         return err;
2629 }
2630
2631 /**
2632  * i40e_vsi_free_tx_resources - Free Tx resources for VSI queues
2633  * @vsi: ptr to the VSI
2634  *
2635  * Free VSI's transmit software resources
2636  **/
2637 static void i40e_vsi_free_tx_resources(struct i40e_vsi *vsi)
2638 {
2639         int i;
2640
2641         if (!vsi->tx_rings)
2642                 return;
2643
2644         for (i = 0; i < vsi->num_queue_pairs; i++)
2645                 if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc)
2646                         i40e_free_tx_resources(vsi->tx_rings[i]);
2647 }
2648
2649 /**
2650  * i40e_vsi_setup_rx_resources - Allocate VSI queues Rx resources
2651  * @vsi: ptr to the VSI
2652  *
2653  * If this function returns with an error, then it's possible one or
2654  * more of the rings is populated (while the rest are not).  It is the
2655  * callers duty to clean those orphaned rings.
2656  *
2657  * Return 0 on success, negative on failure
2658  **/
2659 static int i40e_vsi_setup_rx_resources(struct i40e_vsi *vsi)
2660 {
2661         int i, err = 0;
2662
2663         for (i = 0; i < vsi->num_queue_pairs && !err; i++)
2664                 err = i40e_setup_rx_descriptors(vsi->rx_rings[i]);
2665 #ifdef I40E_FCOE
2666         i40e_fcoe_setup_ddp_resources(vsi);
2667 #endif
2668         return err;
2669 }
2670
2671 /**
2672  * i40e_vsi_free_rx_resources - Free Rx Resources for VSI queues
2673  * @vsi: ptr to the VSI
2674  *
2675  * Free all receive software resources
2676  **/
2677 static void i40e_vsi_free_rx_resources(struct i40e_vsi *vsi)
2678 {
2679         int i;
2680
2681         if (!vsi->rx_rings)
2682                 return;
2683
2684         for (i = 0; i < vsi->num_queue_pairs; i++)
2685                 if (vsi->rx_rings[i] && vsi->rx_rings[i]->desc)
2686                         i40e_free_rx_resources(vsi->rx_rings[i]);
2687 #ifdef I40E_FCOE
2688         i40e_fcoe_free_ddp_resources(vsi);
2689 #endif
2690 }
2691
2692 /**
2693  * i40e_config_xps_tx_ring - Configure XPS for a Tx ring
2694  * @ring: The Tx ring to configure
2695  *
2696  * This enables/disables XPS for a given Tx descriptor ring
2697  * based on the TCs enabled for the VSI that ring belongs to.
2698  **/
2699 static void i40e_config_xps_tx_ring(struct i40e_ring *ring)
2700 {
2701         struct i40e_vsi *vsi = ring->vsi;
2702         cpumask_var_t mask;
2703
2704         if (!ring->q_vector || !ring->netdev)
2705                 return;
2706
2707         /* Single TC mode enable XPS */
2708         if (vsi->tc_config.numtc <= 1) {
2709                 if (!test_and_set_bit(__I40E_TX_XPS_INIT_DONE, &ring->state))
2710                         netif_set_xps_queue(ring->netdev,
2711                                             &ring->q_vector->affinity_mask,
2712                                             ring->queue_index);
2713         } else if (alloc_cpumask_var(&mask, GFP_KERNEL)) {
2714                 /* Disable XPS to allow selection based on TC */
2715                 bitmap_zero(cpumask_bits(mask), nr_cpumask_bits);
2716                 netif_set_xps_queue(ring->netdev, mask, ring->queue_index);
2717                 free_cpumask_var(mask);
2718         }
2719
2720         /* schedule our worker thread which will take care of
2721          * applying the new filter changes
2722          */
2723         i40e_service_event_schedule(vsi->back);
2724 }
2725
2726 /**
2727  * i40e_configure_tx_ring - Configure a transmit ring context and rest
2728  * @ring: The Tx ring to configure
2729  *
2730  * Configure the Tx descriptor ring in the HMC context.
2731  **/
2732 static int i40e_configure_tx_ring(struct i40e_ring *ring)
2733 {
2734         struct i40e_vsi *vsi = ring->vsi;
2735         u16 pf_q = vsi->base_queue + ring->queue_index;
2736         struct i40e_hw *hw = &vsi->back->hw;
2737         struct i40e_hmc_obj_txq tx_ctx;
2738         i40e_status err = 0;
2739         u32 qtx_ctl = 0;
2740
2741         /* some ATR related tx ring init */
2742         if (vsi->back->flags & I40E_FLAG_FD_ATR_ENABLED) {
2743                 ring->atr_sample_rate = vsi->back->atr_sample_rate;
2744                 ring->atr_count = 0;
2745         } else {
2746                 ring->atr_sample_rate = 0;
2747         }
2748
2749         /* configure XPS */
2750         i40e_config_xps_tx_ring(ring);
2751
2752         /* clear the context structure first */
2753         memset(&tx_ctx, 0, sizeof(tx_ctx));
2754
2755         tx_ctx.new_context = 1;
2756         tx_ctx.base = (ring->dma / 128);
2757         tx_ctx.qlen = ring->count;
2758         tx_ctx.fd_ena = !!(vsi->back->flags & (I40E_FLAG_FD_SB_ENABLED |
2759                                                I40E_FLAG_FD_ATR_ENABLED));
2760 #ifdef I40E_FCOE
2761         tx_ctx.fc_ena = (vsi->type == I40E_VSI_FCOE);
2762 #endif
2763         tx_ctx.timesync_ena = !!(vsi->back->flags & I40E_FLAG_PTP);
2764         /* FDIR VSI tx ring can still use RS bit and writebacks */
2765         if (vsi->type != I40E_VSI_FDIR)
2766                 tx_ctx.head_wb_ena = 1;
2767         tx_ctx.head_wb_addr = ring->dma +
2768                               (ring->count * sizeof(struct i40e_tx_desc));
2769
2770         /* As part of VSI creation/update, FW allocates certain
2771          * Tx arbitration queue sets for each TC enabled for
2772          * the VSI. The FW returns the handles to these queue
2773          * sets as part of the response buffer to Add VSI,
2774          * Update VSI, etc. AQ commands. It is expected that
2775          * these queue set handles be associated with the Tx
2776          * queues by the driver as part of the TX queue context
2777          * initialization. This has to be done regardless of
2778          * DCB as by default everything is mapped to TC0.
2779          */
2780         tx_ctx.rdylist = le16_to_cpu(vsi->info.qs_handle[ring->dcb_tc]);
2781         tx_ctx.rdylist_act = 0;
2782
2783         /* clear the context in the HMC */
2784         err = i40e_clear_lan_tx_queue_context(hw, pf_q);
2785         if (err) {
2786                 dev_info(&vsi->back->pdev->dev,
2787                          "Failed to clear LAN Tx queue context on Tx ring %d (pf_q %d), error: %d\n",
2788                          ring->queue_index, pf_q, err);
2789                 return -ENOMEM;
2790         }
2791
2792         /* set the context in the HMC */
2793         err = i40e_set_lan_tx_queue_context(hw, pf_q, &tx_ctx);
2794         if (err) {
2795                 dev_info(&vsi->back->pdev->dev,
2796                          "Failed to set LAN Tx queue context on Tx ring %d (pf_q %d, error: %d\n",
2797                          ring->queue_index, pf_q, err);
2798                 return -ENOMEM;
2799         }
2800
2801         /* Now associate this queue with this PCI function */
2802         if (vsi->type == I40E_VSI_VMDQ2) {
2803                 qtx_ctl = I40E_QTX_CTL_VM_QUEUE;
2804                 qtx_ctl |= ((vsi->id) << I40E_QTX_CTL_VFVM_INDX_SHIFT) &
2805                            I40E_QTX_CTL_VFVM_INDX_MASK;
2806         } else {
2807                 qtx_ctl = I40E_QTX_CTL_PF_QUEUE;
2808         }
2809
2810         qtx_ctl |= ((hw->pf_id << I40E_QTX_CTL_PF_INDX_SHIFT) &
2811                     I40E_QTX_CTL_PF_INDX_MASK);
2812         wr32(hw, I40E_QTX_CTL(pf_q), qtx_ctl);
2813         i40e_flush(hw);
2814
2815         /* cache tail off for easier writes later */
2816         ring->tail = hw->hw_addr + I40E_QTX_TAIL(pf_q);
2817
2818         return 0;
2819 }
2820
2821 /**
2822  * i40e_configure_rx_ring - Configure a receive ring context
2823  * @ring: The Rx ring to configure
2824  *
2825  * Configure the Rx descriptor ring in the HMC context.
2826  **/
2827 static int i40e_configure_rx_ring(struct i40e_ring *ring)
2828 {
2829         struct i40e_vsi *vsi = ring->vsi;
2830         u32 chain_len = vsi->back->hw.func_caps.rx_buf_chain_len;
2831         u16 pf_q = vsi->base_queue + ring->queue_index;
2832         struct i40e_hw *hw = &vsi->back->hw;
2833         struct i40e_hmc_obj_rxq rx_ctx;
2834         i40e_status err = 0;
2835
2836         ring->state = 0;
2837
2838         /* clear the context structure first */
2839         memset(&rx_ctx, 0, sizeof(rx_ctx));
2840
2841         ring->rx_buf_len = vsi->rx_buf_len;
2842         ring->rx_hdr_len = vsi->rx_hdr_len;
2843
2844         rx_ctx.dbuff = ring->rx_buf_len >> I40E_RXQ_CTX_DBUFF_SHIFT;
2845         rx_ctx.hbuff = ring->rx_hdr_len >> I40E_RXQ_CTX_HBUFF_SHIFT;
2846
2847         rx_ctx.base = (ring->dma / 128);
2848         rx_ctx.qlen = ring->count;
2849
2850         if (vsi->back->flags & I40E_FLAG_16BYTE_RX_DESC_ENABLED) {
2851                 set_ring_16byte_desc_enabled(ring);
2852                 rx_ctx.dsize = 0;
2853         } else {
2854                 rx_ctx.dsize = 1;
2855         }
2856
2857         rx_ctx.dtype = vsi->dtype;
2858         if (vsi->dtype) {
2859                 set_ring_ps_enabled(ring);
2860                 rx_ctx.hsplit_0 = I40E_RX_SPLIT_L2      |
2861                                   I40E_RX_SPLIT_IP      |
2862                                   I40E_RX_SPLIT_TCP_UDP |
2863                                   I40E_RX_SPLIT_SCTP;
2864         } else {
2865                 rx_ctx.hsplit_0 = 0;
2866         }
2867
2868         rx_ctx.rxmax = min_t(u16, vsi->max_frame,
2869                                   (chain_len * ring->rx_buf_len));
2870         if (hw->revision_id == 0)
2871                 rx_ctx.lrxqthresh = 0;
2872         else
2873                 rx_ctx.lrxqthresh = 2;
2874         rx_ctx.crcstrip = 1;
2875         rx_ctx.l2tsel = 1;
2876         /* this controls whether VLAN is stripped from inner headers */
2877         rx_ctx.showiv = 0;
2878 #ifdef I40E_FCOE
2879         rx_ctx.fc_ena = (vsi->type == I40E_VSI_FCOE);
2880 #endif
2881         /* set the prefena field to 1 because the manual says to */
2882         rx_ctx.prefena = 1;
2883
2884         /* clear the context in the HMC */
2885         err = i40e_clear_lan_rx_queue_context(hw, pf_q);
2886         if (err) {
2887                 dev_info(&vsi->back->pdev->dev,
2888                          "Failed to clear LAN Rx queue context on Rx ring %d (pf_q %d), error: %d\n",
2889                          ring->queue_index, pf_q, err);
2890                 return -ENOMEM;
2891         }
2892
2893         /* set the context in the HMC */
2894         err = i40e_set_lan_rx_queue_context(hw, pf_q, &rx_ctx);
2895         if (err) {
2896                 dev_info(&vsi->back->pdev->dev,
2897                          "Failed to set LAN Rx queue context on Rx ring %d (pf_q %d), error: %d\n",
2898                          ring->queue_index, pf_q, err);
2899                 return -ENOMEM;
2900         }
2901
2902         /* cache tail for quicker writes, and clear the reg before use */
2903         ring->tail = hw->hw_addr + I40E_QRX_TAIL(pf_q);
2904         writel(0, ring->tail);
2905
2906         if (ring_is_ps_enabled(ring)) {
2907                 i40e_alloc_rx_headers(ring);
2908                 i40e_alloc_rx_buffers_ps(ring, I40E_DESC_UNUSED(ring));
2909         } else {
2910                 i40e_alloc_rx_buffers_1buf(ring, I40E_DESC_UNUSED(ring));
2911         }
2912
2913         return 0;
2914 }
2915
2916 /**
2917  * i40e_vsi_configure_tx - Configure the VSI for Tx
2918  * @vsi: VSI structure describing this set of rings and resources
2919  *
2920  * Configure the Tx VSI for operation.
2921  **/
2922 static int i40e_vsi_configure_tx(struct i40e_vsi *vsi)
2923 {
2924         int err = 0;
2925         u16 i;
2926
2927         for (i = 0; (i < vsi->num_queue_pairs) && !err; i++)
2928                 err = i40e_configure_tx_ring(vsi->tx_rings[i]);
2929
2930         return err;
2931 }
2932
2933 /**
2934  * i40e_vsi_configure_rx - Configure the VSI for Rx
2935  * @vsi: the VSI being configured
2936  *
2937  * Configure the Rx VSI for operation.
2938  **/
2939 static int i40e_vsi_configure_rx(struct i40e_vsi *vsi)
2940 {
2941         int err = 0;
2942         u16 i;
2943
2944         if (vsi->netdev && (vsi->netdev->mtu > ETH_DATA_LEN))
2945                 vsi->max_frame = vsi->netdev->mtu + ETH_HLEN
2946                                + ETH_FCS_LEN + VLAN_HLEN;
2947         else
2948                 vsi->max_frame = I40E_RXBUFFER_2048;
2949
2950         /* figure out correct receive buffer length */
2951         switch (vsi->back->flags & (I40E_FLAG_RX_1BUF_ENABLED |
2952                                     I40E_FLAG_RX_PS_ENABLED)) {
2953         case I40E_FLAG_RX_1BUF_ENABLED:
2954                 vsi->rx_hdr_len = 0;
2955                 vsi->rx_buf_len = vsi->max_frame;
2956                 vsi->dtype = I40E_RX_DTYPE_NO_SPLIT;
2957                 break;
2958         case I40E_FLAG_RX_PS_ENABLED:
2959                 vsi->rx_hdr_len = I40E_RX_HDR_SIZE;
2960                 vsi->rx_buf_len = I40E_RXBUFFER_2048;
2961                 vsi->dtype = I40E_RX_DTYPE_HEADER_SPLIT;
2962                 break;
2963         default:
2964                 vsi->rx_hdr_len = I40E_RX_HDR_SIZE;
2965                 vsi->rx_buf_len = I40E_RXBUFFER_2048;
2966                 vsi->dtype = I40E_RX_DTYPE_SPLIT_ALWAYS;
2967                 break;
2968         }
2969
2970 #ifdef I40E_FCOE
2971         /* setup rx buffer for FCoE */
2972         if ((vsi->type == I40E_VSI_FCOE) &&
2973             (vsi->back->flags & I40E_FLAG_FCOE_ENABLED)) {
2974                 vsi->rx_hdr_len = 0;
2975                 vsi->rx_buf_len = I40E_RXBUFFER_3072;
2976                 vsi->max_frame = I40E_RXBUFFER_3072;
2977                 vsi->dtype = I40E_RX_DTYPE_NO_SPLIT;
2978         }
2979
2980 #endif /* I40E_FCOE */
2981         /* round up for the chip's needs */
2982         vsi->rx_hdr_len = ALIGN(vsi->rx_hdr_len,
2983                                 BIT_ULL(I40E_RXQ_CTX_HBUFF_SHIFT));
2984         vsi->rx_buf_len = ALIGN(vsi->rx_buf_len,
2985                                 BIT_ULL(I40E_RXQ_CTX_DBUFF_SHIFT));
2986
2987         /* set up individual rings */
2988         for (i = 0; i < vsi->num_queue_pairs && !err; i++)
2989                 err = i40e_configure_rx_ring(vsi->rx_rings[i]);
2990
2991         return err;
2992 }
2993
2994 /**
2995  * i40e_vsi_config_dcb_rings - Update rings to reflect DCB TC
2996  * @vsi: ptr to the VSI
2997  **/
2998 static void i40e_vsi_config_dcb_rings(struct i40e_vsi *vsi)
2999 {
3000         struct i40e_ring *tx_ring, *rx_ring;
3001         u16 qoffset, qcount;
3002         int i, n;
3003
3004         if (!(vsi->back->flags & I40E_FLAG_DCB_ENABLED)) {
3005                 /* Reset the TC information */
3006                 for (i = 0; i < vsi->num_queue_pairs; i++) {
3007                         rx_ring = vsi->rx_rings[i];
3008                         tx_ring = vsi->tx_rings[i];
3009                         rx_ring->dcb_tc = 0;
3010                         tx_ring->dcb_tc = 0;
3011                 }
3012         }
3013
3014         for (n = 0; n < I40E_MAX_TRAFFIC_CLASS; n++) {
3015                 if (!(vsi->tc_config.enabled_tc & BIT_ULL(n)))
3016                         continue;
3017
3018                 qoffset = vsi->tc_config.tc_info[n].qoffset;
3019                 qcount = vsi->tc_config.tc_info[n].qcount;
3020                 for (i = qoffset; i < (qoffset + qcount); i++) {
3021                         rx_ring = vsi->rx_rings[i];
3022                         tx_ring = vsi->tx_rings[i];
3023                         rx_ring->dcb_tc = n;
3024                         tx_ring->dcb_tc = n;
3025                 }
3026         }
3027 }
3028
3029 /**
3030  * i40e_set_vsi_rx_mode - Call set_rx_mode on a VSI
3031  * @vsi: ptr to the VSI
3032  **/
3033 static void i40e_set_vsi_rx_mode(struct i40e_vsi *vsi)
3034 {
3035         if (vsi->netdev)
3036                 i40e_set_rx_mode(vsi->netdev);
3037 }
3038
3039 /**
3040  * i40e_fdir_filter_restore - Restore the Sideband Flow Director filters
3041  * @vsi: Pointer to the targeted VSI
3042  *
3043  * This function replays the hlist on the hw where all the SB Flow Director
3044  * filters were saved.
3045  **/
3046 static void i40e_fdir_filter_restore(struct i40e_vsi *vsi)
3047 {
3048         struct i40e_fdir_filter *filter;
3049         struct i40e_pf *pf = vsi->back;
3050         struct hlist_node *node;
3051
3052         if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
3053                 return;
3054
3055         hlist_for_each_entry_safe(filter, node,
3056                                   &pf->fdir_filter_list, fdir_node) {
3057                 i40e_add_del_fdir(vsi, filter, true);
3058         }
3059 }
3060
3061 /**
3062  * i40e_vsi_configure - Set up the VSI for action
3063  * @vsi: the VSI being configured
3064  **/
3065 static int i40e_vsi_configure(struct i40e_vsi *vsi)
3066 {
3067         int err;
3068
3069         i40e_set_vsi_rx_mode(vsi);
3070         i40e_restore_vlan(vsi);
3071         i40e_vsi_config_dcb_rings(vsi);
3072         err = i40e_vsi_configure_tx(vsi);
3073         if (!err)
3074                 err = i40e_vsi_configure_rx(vsi);
3075
3076         return err;
3077 }
3078
3079 /**
3080  * i40e_vsi_configure_msix - MSIX mode Interrupt Config in the HW
3081  * @vsi: the VSI being configured
3082  **/
3083 static void i40e_vsi_configure_msix(struct i40e_vsi *vsi)
3084 {
3085         struct i40e_pf *pf = vsi->back;
3086         struct i40e_hw *hw = &pf->hw;
3087         u16 vector;
3088         int i, q;
3089         u32 qp;
3090
3091         /* The interrupt indexing is offset by 1 in the PFINT_ITRn
3092          * and PFINT_LNKLSTn registers, e.g.:
3093          *   PFINT_ITRn[0..n-1] gets msix-1..msix-n  (qpair interrupts)
3094          */
3095         qp = vsi->base_queue;
3096         vector = vsi->base_vector;
3097         for (i = 0; i < vsi->num_q_vectors; i++, vector++) {
3098                 struct i40e_q_vector *q_vector = vsi->q_vectors[i];
3099
3100                 q_vector->itr_countdown = ITR_COUNTDOWN_START;
3101                 q_vector->rx.itr = ITR_TO_REG(vsi->rx_itr_setting);
3102                 q_vector->rx.latency_range = I40E_LOW_LATENCY;
3103                 wr32(hw, I40E_PFINT_ITRN(I40E_RX_ITR, vector - 1),
3104                      q_vector->rx.itr);
3105                 q_vector->tx.itr = ITR_TO_REG(vsi->tx_itr_setting);
3106                 q_vector->tx.latency_range = I40E_LOW_LATENCY;
3107                 wr32(hw, I40E_PFINT_ITRN(I40E_TX_ITR, vector - 1),
3108                      q_vector->tx.itr);
3109                 wr32(hw, I40E_PFINT_RATEN(vector - 1),
3110                      INTRL_USEC_TO_REG(vsi->int_rate_limit));
3111
3112                 /* Linked list for the queuepairs assigned to this vector */
3113                 wr32(hw, I40E_PFINT_LNKLSTN(vector - 1), qp);
3114                 for (q = 0; q < q_vector->num_ringpairs; q++) {
3115                         u32 val;
3116
3117                         val = I40E_QINT_RQCTL_CAUSE_ENA_MASK |
3118                               (I40E_RX_ITR << I40E_QINT_RQCTL_ITR_INDX_SHIFT)  |
3119                               (vector      << I40E_QINT_RQCTL_MSIX_INDX_SHIFT) |
3120                               (qp          << I40E_QINT_RQCTL_NEXTQ_INDX_SHIFT)|
3121                               (I40E_QUEUE_TYPE_TX
3122                                       << I40E_QINT_RQCTL_NEXTQ_TYPE_SHIFT);
3123
3124                         wr32(hw, I40E_QINT_RQCTL(qp), val);
3125
3126                         val = I40E_QINT_TQCTL_CAUSE_ENA_MASK |
3127                               (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT)  |
3128                               (vector      << I40E_QINT_TQCTL_MSIX_INDX_SHIFT) |
3129                               ((qp+1)      << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT)|
3130                               (I40E_QUEUE_TYPE_RX
3131                                       << I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
3132
3133                         /* Terminate the linked list */
3134                         if (q == (q_vector->num_ringpairs - 1))
3135                                 val |= (I40E_QUEUE_END_OF_LIST
3136                                            << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT);
3137
3138                         wr32(hw, I40E_QINT_TQCTL(qp), val);
3139                         qp++;
3140                 }
3141         }
3142
3143         i40e_flush(hw);
3144 }
3145
3146 /**
3147  * i40e_enable_misc_int_causes - enable the non-queue interrupts
3148  * @hw: ptr to the hardware info
3149  **/
3150 static void i40e_enable_misc_int_causes(struct i40e_pf *pf)
3151 {
3152         struct i40e_hw *hw = &pf->hw;
3153         u32 val;
3154
3155         /* clear things first */
3156         wr32(hw, I40E_PFINT_ICR0_ENA, 0);  /* disable all */
3157         rd32(hw, I40E_PFINT_ICR0);         /* read to clear */
3158
3159         val = I40E_PFINT_ICR0_ENA_ECC_ERR_MASK       |
3160               I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK    |
3161               I40E_PFINT_ICR0_ENA_GRST_MASK          |
3162               I40E_PFINT_ICR0_ENA_PCI_EXCEPTION_MASK |
3163               I40E_PFINT_ICR0_ENA_GPIO_MASK          |
3164               I40E_PFINT_ICR0_ENA_HMC_ERR_MASK       |
3165               I40E_PFINT_ICR0_ENA_VFLR_MASK          |
3166               I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
3167
3168         if (pf->flags & I40E_FLAG_IWARP_ENABLED)
3169                 val |= I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK;
3170
3171         if (pf->flags & I40E_FLAG_PTP)
3172                 val |= I40E_PFINT_ICR0_ENA_TIMESYNC_MASK;
3173
3174         wr32(hw, I40E_PFINT_ICR0_ENA, val);
3175
3176         /* SW_ITR_IDX = 0, but don't change INTENA */
3177         wr32(hw, I40E_PFINT_DYN_CTL0, I40E_PFINT_DYN_CTL0_SW_ITR_INDX_MASK |
3178                                         I40E_PFINT_DYN_CTL0_INTENA_MSK_MASK);
3179
3180         /* OTHER_ITR_IDX = 0 */
3181         wr32(hw, I40E_PFINT_STAT_CTL0, 0);
3182 }
3183
3184 /**
3185  * i40e_configure_msi_and_legacy - Legacy mode interrupt config in the HW
3186  * @vsi: the VSI being configured
3187  **/
3188 static void i40e_configure_msi_and_legacy(struct i40e_vsi *vsi)
3189 {
3190         struct i40e_q_vector *q_vector = vsi->q_vectors[0];
3191         struct i40e_pf *pf = vsi->back;
3192         struct i40e_hw *hw = &pf->hw;
3193         u32 val;
3194
3195         /* set the ITR configuration */
3196         q_vector->itr_countdown = ITR_COUNTDOWN_START;
3197         q_vector->rx.itr = ITR_TO_REG(vsi->rx_itr_setting);
3198         q_vector->rx.latency_range = I40E_LOW_LATENCY;
3199         wr32(hw, I40E_PFINT_ITR0(I40E_RX_ITR), q_vector->rx.itr);
3200         q_vector->tx.itr = ITR_TO_REG(vsi->tx_itr_setting);
3201         q_vector->tx.latency_range = I40E_LOW_LATENCY;
3202         wr32(hw, I40E_PFINT_ITR0(I40E_TX_ITR), q_vector->tx.itr);
3203
3204         i40e_enable_misc_int_causes(pf);
3205
3206         /* FIRSTQ_INDX = 0, FIRSTQ_TYPE = 0 (rx) */
3207         wr32(hw, I40E_PFINT_LNKLST0, 0);
3208
3209         /* Associate the queue pair to the vector and enable the queue int */
3210         val = I40E_QINT_RQCTL_CAUSE_ENA_MASK                  |
3211               (I40E_RX_ITR << I40E_QINT_RQCTL_ITR_INDX_SHIFT) |
3212               (I40E_QUEUE_TYPE_TX << I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
3213
3214         wr32(hw, I40E_QINT_RQCTL(0), val);
3215
3216         val = I40E_QINT_TQCTL_CAUSE_ENA_MASK                  |
3217               (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT) |
3218               (I40E_QUEUE_END_OF_LIST << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT);
3219
3220         wr32(hw, I40E_QINT_TQCTL(0), val);
3221         i40e_flush(hw);
3222 }
3223
3224 /**
3225  * i40e_irq_dynamic_disable_icr0 - Disable default interrupt generation for icr0
3226  * @pf: board private structure
3227  **/
3228 void i40e_irq_dynamic_disable_icr0(struct i40e_pf *pf)
3229 {
3230         struct i40e_hw *hw = &pf->hw;
3231
3232         wr32(hw, I40E_PFINT_DYN_CTL0,
3233              I40E_ITR_NONE << I40E_PFINT_DYN_CTLN_ITR_INDX_SHIFT);
3234         i40e_flush(hw);
3235 }
3236
3237 /**
3238  * i40e_irq_dynamic_enable_icr0 - Enable default interrupt generation for icr0
3239  * @pf: board private structure
3240  **/
3241 void i40e_irq_dynamic_enable_icr0(struct i40e_pf *pf)
3242 {
3243         struct i40e_hw *hw = &pf->hw;
3244         u32 val;
3245
3246         val = I40E_PFINT_DYN_CTL0_INTENA_MASK   |
3247               I40E_PFINT_DYN_CTL0_CLEARPBA_MASK |
3248               (I40E_ITR_NONE << I40E_PFINT_DYN_CTL0_ITR_INDX_SHIFT);
3249
3250         wr32(hw, I40E_PFINT_DYN_CTL0, val);
3251         i40e_flush(hw);
3252 }
3253
3254 /**
3255  * i40e_irq_dynamic_disable - Disable default interrupt generation settings
3256  * @vsi: pointer to a vsi
3257  * @vector: disable a particular Hw Interrupt vector
3258  **/
3259 void i40e_irq_dynamic_disable(struct i40e_vsi *vsi, int vector)
3260 {
3261         struct i40e_pf *pf = vsi->back;
3262         struct i40e_hw *hw = &pf->hw;
3263         u32 val;
3264
3265         val = I40E_ITR_NONE << I40E_PFINT_DYN_CTLN_ITR_INDX_SHIFT;
3266         wr32(hw, I40E_PFINT_DYN_CTLN(vector - 1), val);
3267         i40e_flush(hw);
3268 }
3269
3270 /**
3271  * i40e_msix_clean_rings - MSIX mode Interrupt Handler
3272  * @irq: interrupt number
3273  * @data: pointer to a q_vector
3274  **/
3275 static irqreturn_t i40e_msix_clean_rings(int irq, void *data)
3276 {
3277         struct i40e_q_vector *q_vector = data;
3278
3279         if (!q_vector->tx.ring && !q_vector->rx.ring)
3280                 return IRQ_HANDLED;
3281
3282         napi_schedule_irqoff(&q_vector->napi);
3283
3284         return IRQ_HANDLED;
3285 }
3286
3287 /**
3288  * i40e_vsi_request_irq_msix - Initialize MSI-X interrupts
3289  * @vsi: the VSI being configured
3290  * @basename: name for the vector
3291  *
3292  * Allocates MSI-X vectors and requests interrupts from the kernel.
3293  **/
3294 static int i40e_vsi_request_irq_msix(struct i40e_vsi *vsi, char *basename)
3295 {
3296         int q_vectors = vsi->num_q_vectors;
3297         struct i40e_pf *pf = vsi->back;
3298         int base = vsi->base_vector;
3299         int rx_int_idx = 0;
3300         int tx_int_idx = 0;
3301         int vector, err;
3302
3303         for (vector = 0; vector < q_vectors; vector++) {
3304                 struct i40e_q_vector *q_vector = vsi->q_vectors[vector];
3305
3306                 if (q_vector->tx.ring && q_vector->rx.ring) {
3307                         snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3308                                  "%s-%s-%d", basename, "TxRx", rx_int_idx++);
3309                         tx_int_idx++;
3310                 } else if (q_vector->rx.ring) {
3311                         snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3312                                  "%s-%s-%d", basename, "rx", rx_int_idx++);
3313                 } else if (q_vector->tx.ring) {
3314                         snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3315                                  "%s-%s-%d", basename, "tx", tx_int_idx++);
3316                 } else {
3317                         /* skip this unused q_vector */
3318                         continue;
3319                 }
3320                 err = request_irq(pf->msix_entries[base + vector].vector,
3321                                   vsi->irq_handler,
3322                                   0,
3323                                   q_vector->name,
3324                                   q_vector);
3325                 if (err) {
3326                         dev_info(&pf->pdev->dev,
3327                                  "MSIX request_irq failed, error: %d\n", err);
3328                         goto free_queue_irqs;
3329                 }
3330                 /* assign the mask for this irq */
3331                 irq_set_affinity_hint(pf->msix_entries[base + vector].vector,
3332                                       &q_vector->affinity_mask);
3333         }
3334
3335         vsi->irqs_ready = true;
3336         return 0;
3337
3338 free_queue_irqs:
3339         while (vector) {
3340                 vector--;
3341                 irq_set_affinity_hint(pf->msix_entries[base + vector].vector,
3342                                       NULL);
3343                 free_irq(pf->msix_entries[base + vector].vector,
3344                          &(vsi->q_vectors[vector]));
3345         }
3346         return err;
3347 }
3348
3349 /**
3350  * i40e_vsi_disable_irq - Mask off queue interrupt generation on the VSI
3351  * @vsi: the VSI being un-configured
3352  **/
3353 static void i40e_vsi_disable_irq(struct i40e_vsi *vsi)
3354 {
3355         struct i40e_pf *pf = vsi->back;
3356         struct i40e_hw *hw = &pf->hw;
3357         int base = vsi->base_vector;
3358         int i;
3359
3360         for (i = 0; i < vsi->num_queue_pairs; i++) {
3361                 wr32(hw, I40E_QINT_TQCTL(vsi->tx_rings[i]->reg_idx), 0);
3362                 wr32(hw, I40E_QINT_RQCTL(vsi->rx_rings[i]->reg_idx), 0);
3363         }
3364
3365         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3366                 for (i = vsi->base_vector;
3367                      i < (vsi->num_q_vectors + vsi->base_vector); i++)
3368                         wr32(hw, I40E_PFINT_DYN_CTLN(i - 1), 0);
3369
3370                 i40e_flush(hw);
3371                 for (i = 0; i < vsi->num_q_vectors; i++)
3372                         synchronize_irq(pf->msix_entries[i + base].vector);
3373         } else {
3374                 /* Legacy and MSI mode - this stops all interrupt handling */
3375                 wr32(hw, I40E_PFINT_ICR0_ENA, 0);
3376                 wr32(hw, I40E_PFINT_DYN_CTL0, 0);
3377                 i40e_flush(hw);
3378                 synchronize_irq(pf->pdev->irq);
3379         }
3380 }
3381
3382 /**
3383  * i40e_vsi_enable_irq - Enable IRQ for the given VSI
3384  * @vsi: the VSI being configured
3385  **/
3386 static int i40e_vsi_enable_irq(struct i40e_vsi *vsi)
3387 {
3388         struct i40e_pf *pf = vsi->back;
3389         int i;
3390
3391         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3392                 for (i = 0; i < vsi->num_q_vectors; i++)
3393                         i40e_irq_dynamic_enable(vsi, i);
3394         } else {
3395                 i40e_irq_dynamic_enable_icr0(pf);
3396         }
3397
3398         i40e_flush(&pf->hw);
3399         return 0;
3400 }
3401
3402 /**
3403  * i40e_stop_misc_vector - Stop the vector that handles non-queue events
3404  * @pf: board private structure
3405  **/
3406 static void i40e_stop_misc_vector(struct i40e_pf *pf)
3407 {
3408         /* Disable ICR 0 */
3409         wr32(&pf->hw, I40E_PFINT_ICR0_ENA, 0);
3410         i40e_flush(&pf->hw);
3411 }
3412
3413 /**
3414  * i40e_intr - MSI/Legacy and non-queue interrupt handler
3415  * @irq: interrupt number
3416  * @data: pointer to a q_vector
3417  *
3418  * This is the handler used for all MSI/Legacy interrupts, and deals
3419  * with both queue and non-queue interrupts.  This is also used in
3420  * MSIX mode to handle the non-queue interrupts.
3421  **/
3422 static irqreturn_t i40e_intr(int irq, void *data)
3423 {
3424         struct i40e_pf *pf = (struct i40e_pf *)data;
3425         struct i40e_hw *hw = &pf->hw;
3426         irqreturn_t ret = IRQ_NONE;
3427         u32 icr0, icr0_remaining;
3428         u32 val, ena_mask;
3429
3430         icr0 = rd32(hw, I40E_PFINT_ICR0);
3431         ena_mask = rd32(hw, I40E_PFINT_ICR0_ENA);
3432
3433         /* if sharing a legacy IRQ, we might get called w/o an intr pending */
3434         if ((icr0 & I40E_PFINT_ICR0_INTEVENT_MASK) == 0)
3435                 goto enable_intr;
3436
3437         /* if interrupt but no bits showing, must be SWINT */
3438         if (((icr0 & ~I40E_PFINT_ICR0_INTEVENT_MASK) == 0) ||
3439             (icr0 & I40E_PFINT_ICR0_SWINT_MASK))
3440                 pf->sw_int_count++;
3441
3442         if ((pf->flags & I40E_FLAG_IWARP_ENABLED) &&
3443             (ena_mask & I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK)) {
3444                 ena_mask &= ~I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK;
3445                 icr0 &= ~I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK;
3446                 dev_info(&pf->pdev->dev, "cleared PE_CRITERR\n");
3447         }
3448
3449         /* only q0 is used in MSI/Legacy mode, and none are used in MSIX */
3450         if (icr0 & I40E_PFINT_ICR0_QUEUE_0_MASK) {
3451                 struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
3452                 struct i40e_q_vector *q_vector = vsi->q_vectors[0];
3453
3454                 /* temporarily disable queue cause for NAPI processing */
3455                 u32 qval = rd32(hw, I40E_QINT_RQCTL(0));
3456
3457                 qval &= ~I40E_QINT_RQCTL_CAUSE_ENA_MASK;
3458                 wr32(hw, I40E_QINT_RQCTL(0), qval);
3459
3460                 qval = rd32(hw, I40E_QINT_TQCTL(0));
3461                 qval &= ~I40E_QINT_TQCTL_CAUSE_ENA_MASK;
3462                 wr32(hw, I40E_QINT_TQCTL(0), qval);
3463
3464                 if (!test_bit(__I40E_DOWN, &pf->state))
3465                         napi_schedule_irqoff(&q_vector->napi);
3466         }
3467
3468         if (icr0 & I40E_PFINT_ICR0_ADMINQ_MASK) {
3469                 ena_mask &= ~I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
3470                 set_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state);
3471         }
3472
3473         if (icr0 & I40E_PFINT_ICR0_MAL_DETECT_MASK) {
3474                 ena_mask &= ~I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK;
3475                 set_bit(__I40E_MDD_EVENT_PENDING, &pf->state);
3476         }
3477
3478         if (icr0 & I40E_PFINT_ICR0_VFLR_MASK) {
3479                 ena_mask &= ~I40E_PFINT_ICR0_ENA_VFLR_MASK;
3480                 set_bit(__I40E_VFLR_EVENT_PENDING, &pf->state);
3481         }
3482
3483         if (icr0 & I40E_PFINT_ICR0_GRST_MASK) {
3484                 if (!test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state))
3485                         set_bit(__I40E_RESET_INTR_RECEIVED, &pf->state);
3486                 ena_mask &= ~I40E_PFINT_ICR0_ENA_GRST_MASK;
3487                 val = rd32(hw, I40E_GLGEN_RSTAT);
3488                 val = (val & I40E_GLGEN_RSTAT_RESET_TYPE_MASK)
3489                        >> I40E_GLGEN_RSTAT_RESET_TYPE_SHIFT;
3490                 if (val == I40E_RESET_CORER) {
3491                         pf->corer_count++;
3492                 } else if (val == I40E_RESET_GLOBR) {
3493                         pf->globr_count++;
3494                 } else if (val == I40E_RESET_EMPR) {
3495                         pf->empr_count++;
3496                         set_bit(__I40E_EMP_RESET_INTR_RECEIVED, &pf->state);
3497                 }
3498         }
3499
3500         if (icr0 & I40E_PFINT_ICR0_HMC_ERR_MASK) {
3501                 icr0 &= ~I40E_PFINT_ICR0_HMC_ERR_MASK;
3502                 dev_info(&pf->pdev->dev, "HMC error interrupt\n");
3503                 dev_info(&pf->pdev->dev, "HMC error info 0x%x, HMC error data 0x%x\n",
3504                          rd32(hw, I40E_PFHMC_ERRORINFO),
3505                          rd32(hw, I40E_PFHMC_ERRORDATA));
3506         }
3507
3508         if (icr0 & I40E_PFINT_ICR0_TIMESYNC_MASK) {
3509                 u32 prttsyn_stat = rd32(hw, I40E_PRTTSYN_STAT_0);
3510
3511                 if (prttsyn_stat & I40E_PRTTSYN_STAT_0_TXTIME_MASK) {
3512                         icr0 &= ~I40E_PFINT_ICR0_ENA_TIMESYNC_MASK;
3513                         i40e_ptp_tx_hwtstamp(pf);
3514                 }
3515         }
3516
3517         /* If a critical error is pending we have no choice but to reset the
3518          * device.
3519          * Report and mask out any remaining unexpected interrupts.
3520          */
3521         icr0_remaining = icr0 & ena_mask;
3522         if (icr0_remaining) {
3523                 dev_info(&pf->pdev->dev, "unhandled interrupt icr0=0x%08x\n",
3524                          icr0_remaining);
3525                 if ((icr0_remaining & I40E_PFINT_ICR0_PE_CRITERR_MASK) ||
3526                     (icr0_remaining & I40E_PFINT_ICR0_PCI_EXCEPTION_MASK) ||
3527                     (icr0_remaining & I40E_PFINT_ICR0_ECC_ERR_MASK)) {
3528                         dev_info(&pf->pdev->dev, "device will be reset\n");
3529                         set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
3530                         i40e_service_event_schedule(pf);
3531                 }
3532                 ena_mask &= ~icr0_remaining;
3533         }
3534         ret = IRQ_HANDLED;
3535
3536 enable_intr:
3537         /* re-enable interrupt causes */
3538         wr32(hw, I40E_PFINT_ICR0_ENA, ena_mask);
3539         if (!test_bit(__I40E_DOWN, &pf->state)) {
3540                 i40e_service_event_schedule(pf);
3541                 i40e_irq_dynamic_enable_icr0(pf);
3542         }
3543
3544         return ret;
3545 }
3546
3547 /**
3548  * i40e_clean_fdir_tx_irq - Reclaim resources after transmit completes
3549  * @tx_ring:  tx ring to clean
3550  * @budget:   how many cleans we're allowed
3551  *
3552  * Returns true if there's any budget left (e.g. the clean is finished)
3553  **/
3554 static bool i40e_clean_fdir_tx_irq(struct i40e_ring *tx_ring, int budget)
3555 {
3556         struct i40e_vsi *vsi = tx_ring->vsi;
3557         u16 i = tx_ring->next_to_clean;
3558         struct i40e_tx_buffer *tx_buf;
3559         struct i40e_tx_desc *tx_desc;
3560
3561         tx_buf = &tx_ring->tx_bi[i];
3562         tx_desc = I40E_TX_DESC(tx_ring, i);
3563         i -= tx_ring->count;
3564
3565         do {
3566                 struct i40e_tx_desc *eop_desc = tx_buf->next_to_watch;
3567
3568                 /* if next_to_watch is not set then there is no work pending */
3569                 if (!eop_desc)
3570                         break;
3571
3572                 /* prevent any other reads prior to eop_desc */
3573                 read_barrier_depends();
3574
3575                 /* if the descriptor isn't done, no work yet to do */
3576                 if (!(eop_desc->cmd_type_offset_bsz &
3577                       cpu_to_le64(I40E_TX_DESC_DTYPE_DESC_DONE)))
3578                         break;
3579
3580                 /* clear next_to_watch to prevent false hangs */
3581                 tx_buf->next_to_watch = NULL;
3582
3583                 tx_desc->buffer_addr = 0;
3584                 tx_desc->cmd_type_offset_bsz = 0;
3585                 /* move past filter desc */
3586                 tx_buf++;
3587                 tx_desc++;
3588                 i++;
3589                 if (unlikely(!i)) {
3590                         i -= tx_ring->count;
3591                         tx_buf = tx_ring->tx_bi;
3592                         tx_desc = I40E_TX_DESC(tx_ring, 0);
3593                 }
3594                 /* unmap skb header data */
3595                 dma_unmap_single(tx_ring->dev,
3596                                  dma_unmap_addr(tx_buf, dma),
3597                                  dma_unmap_len(tx_buf, len),
3598                                  DMA_TO_DEVICE);
3599                 if (tx_buf->tx_flags & I40E_TX_FLAGS_FD_SB)
3600                         kfree(tx_buf->raw_buf);
3601
3602                 tx_buf->raw_buf = NULL;
3603                 tx_buf->tx_flags = 0;
3604                 tx_buf->next_to_watch = NULL;
3605                 dma_unmap_len_set(tx_buf, len, 0);
3606                 tx_desc->buffer_addr = 0;
3607                 tx_desc->cmd_type_offset_bsz = 0;
3608
3609                 /* move us past the eop_desc for start of next FD desc */
3610                 tx_buf++;
3611                 tx_desc++;
3612                 i++;
3613                 if (unlikely(!i)) {
3614                         i -= tx_ring->count;
3615                         tx_buf = tx_ring->tx_bi;
3616                         tx_desc = I40E_TX_DESC(tx_ring, 0);
3617                 }
3618
3619                 /* update budget accounting */
3620                 budget--;
3621         } while (likely(budget));
3622
3623         i += tx_ring->count;
3624         tx_ring->next_to_clean = i;
3625
3626         if (vsi->back->flags & I40E_FLAG_MSIX_ENABLED)
3627                 i40e_irq_dynamic_enable(vsi, tx_ring->q_vector->v_idx);
3628
3629         return budget > 0;
3630 }
3631
3632 /**
3633  * i40e_fdir_clean_ring - Interrupt Handler for FDIR SB ring
3634  * @irq: interrupt number
3635  * @data: pointer to a q_vector
3636  **/
3637 static irqreturn_t i40e_fdir_clean_ring(int irq, void *data)
3638 {
3639         struct i40e_q_vector *q_vector = data;
3640         struct i40e_vsi *vsi;
3641
3642         if (!q_vector->tx.ring)
3643                 return IRQ_HANDLED;
3644
3645         vsi = q_vector->tx.ring->vsi;
3646         i40e_clean_fdir_tx_irq(q_vector->tx.ring, vsi->work_limit);
3647
3648         return IRQ_HANDLED;
3649 }
3650
3651 /**
3652  * i40e_map_vector_to_qp - Assigns the queue pair to the vector
3653  * @vsi: the VSI being configured
3654  * @v_idx: vector index
3655  * @qp_idx: queue pair index
3656  **/
3657 static void i40e_map_vector_to_qp(struct i40e_vsi *vsi, int v_idx, int qp_idx)
3658 {
3659         struct i40e_q_vector *q_vector = vsi->q_vectors[v_idx];
3660         struct i40e_ring *tx_ring = vsi->tx_rings[qp_idx];
3661         struct i40e_ring *rx_ring = vsi->rx_rings[qp_idx];
3662
3663         tx_ring->q_vector = q_vector;
3664         tx_ring->next = q_vector->tx.ring;
3665         q_vector->tx.ring = tx_ring;
3666         q_vector->tx.count++;
3667
3668         rx_ring->q_vector = q_vector;
3669         rx_ring->next = q_vector->rx.ring;
3670         q_vector->rx.ring = rx_ring;
3671         q_vector->rx.count++;
3672 }
3673
3674 /**
3675  * i40e_vsi_map_rings_to_vectors - Maps descriptor rings to vectors
3676  * @vsi: the VSI being configured
3677  *
3678  * This function maps descriptor rings to the queue-specific vectors
3679  * we were allotted through the MSI-X enabling code.  Ideally, we'd have
3680  * one vector per queue pair, but on a constrained vector budget, we
3681  * group the queue pairs as "efficiently" as possible.
3682  **/
3683 static void i40e_vsi_map_rings_to_vectors(struct i40e_vsi *vsi)
3684 {
3685         int qp_remaining = vsi->num_queue_pairs;
3686         int q_vectors = vsi->num_q_vectors;
3687         int num_ringpairs;
3688         int v_start = 0;
3689         int qp_idx = 0;
3690
3691         /* If we don't have enough vectors for a 1-to-1 mapping, we'll have to
3692          * group them so there are multiple queues per vector.
3693          * It is also important to go through all the vectors available to be
3694          * sure that if we don't use all the vectors, that the remaining vectors
3695          * are cleared. This is especially important when decreasing the
3696          * number of queues in use.
3697          */
3698         for (; v_start < q_vectors; v_start++) {
3699                 struct i40e_q_vector *q_vector = vsi->q_vectors[v_start];
3700
3701                 num_ringpairs = DIV_ROUND_UP(qp_remaining, q_vectors - v_start);
3702
3703                 q_vector->num_ringpairs = num_ringpairs;
3704
3705                 q_vector->rx.count = 0;
3706                 q_vector->tx.count = 0;
3707                 q_vector->rx.ring = NULL;
3708                 q_vector->tx.ring = NULL;
3709
3710                 while (num_ringpairs--) {
3711                         i40e_map_vector_to_qp(vsi, v_start, qp_idx);
3712                         qp_idx++;
3713                         qp_remaining--;
3714                 }
3715         }
3716 }
3717
3718 /**
3719  * i40e_vsi_request_irq - Request IRQ from the OS
3720  * @vsi: the VSI being configured
3721  * @basename: name for the vector
3722  **/
3723 static int i40e_vsi_request_irq(struct i40e_vsi *vsi, char *basename)
3724 {
3725         struct i40e_pf *pf = vsi->back;
3726         int err;
3727
3728         if (pf->flags & I40E_FLAG_MSIX_ENABLED)
3729                 err = i40e_vsi_request_irq_msix(vsi, basename);
3730         else if (pf->flags & I40E_FLAG_MSI_ENABLED)
3731                 err = request_irq(pf->pdev->irq, i40e_intr, 0,
3732                                   pf->int_name, pf);
3733         else
3734                 err = request_irq(pf->pdev->irq, i40e_intr, IRQF_SHARED,
3735                                   pf->int_name, pf);
3736
3737         if (err)
3738                 dev_info(&pf->pdev->dev, "request_irq failed, Error %d\n", err);
3739
3740         return err;
3741 }
3742
3743 #ifdef CONFIG_NET_POLL_CONTROLLER
3744 /**
3745  * i40e_netpoll - A Polling 'interrupt'handler
3746  * @netdev: network interface device structure
3747  *
3748  * This is used by netconsole to send skbs without having to re-enable
3749  * interrupts.  It's not called while the normal interrupt routine is executing.
3750  **/
3751 #ifdef I40E_FCOE
3752 void i40e_netpoll(struct net_device *netdev)
3753 #else
3754 static void i40e_netpoll(struct net_device *netdev)
3755 #endif
3756 {
3757         struct i40e_netdev_priv *np = netdev_priv(netdev);
3758         struct i40e_vsi *vsi = np->vsi;
3759         struct i40e_pf *pf = vsi->back;
3760         int i;
3761
3762         /* if interface is down do nothing */
3763         if (test_bit(__I40E_DOWN, &vsi->state))
3764                 return;
3765
3766         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3767                 for (i = 0; i < vsi->num_q_vectors; i++)
3768                         i40e_msix_clean_rings(0, vsi->q_vectors[i]);
3769         } else {
3770                 i40e_intr(pf->pdev->irq, netdev);
3771         }
3772 }
3773 #endif
3774
3775 /**
3776  * i40e_pf_txq_wait - Wait for a PF's Tx queue to be enabled or disabled
3777  * @pf: the PF being configured
3778  * @pf_q: the PF queue
3779  * @enable: enable or disable state of the queue
3780  *
3781  * This routine will wait for the given Tx queue of the PF to reach the
3782  * enabled or disabled state.
3783  * Returns -ETIMEDOUT in case of failing to reach the requested state after
3784  * multiple retries; else will return 0 in case of success.
3785  **/
3786 static int i40e_pf_txq_wait(struct i40e_pf *pf, int pf_q, bool enable)
3787 {
3788         int i;
3789         u32 tx_reg;
3790
3791         for (i = 0; i < I40E_QUEUE_WAIT_RETRY_LIMIT; i++) {
3792                 tx_reg = rd32(&pf->hw, I40E_QTX_ENA(pf_q));
3793                 if (enable == !!(tx_reg & I40E_QTX_ENA_QENA_STAT_MASK))
3794                         break;
3795
3796                 usleep_range(10, 20);
3797         }
3798         if (i >= I40E_QUEUE_WAIT_RETRY_LIMIT)
3799                 return -ETIMEDOUT;
3800
3801         return 0;
3802 }
3803
3804 /**
3805  * i40e_vsi_control_tx - Start or stop a VSI's rings
3806  * @vsi: the VSI being configured
3807  * @enable: start or stop the rings
3808  **/
3809 static int i40e_vsi_control_tx(struct i40e_vsi *vsi, bool enable)
3810 {
3811         struct i40e_pf *pf = vsi->back;
3812         struct i40e_hw *hw = &pf->hw;
3813         int i, j, pf_q, ret = 0;
3814         u32 tx_reg;
3815
3816         pf_q = vsi->base_queue;
3817         for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
3818
3819                 /* warn the TX unit of coming changes */
3820                 i40e_pre_tx_queue_cfg(&pf->hw, pf_q, enable);
3821                 if (!enable)
3822                         usleep_range(10, 20);
3823
3824                 for (j = 0; j < 50; j++) {
3825                         tx_reg = rd32(hw, I40E_QTX_ENA(pf_q));
3826                         if (((tx_reg >> I40E_QTX_ENA_QENA_REQ_SHIFT) & 1) ==
3827                             ((tx_reg >> I40E_QTX_ENA_QENA_STAT_SHIFT) & 1))
3828                                 break;
3829                         usleep_range(1000, 2000);
3830                 }
3831                 /* Skip if the queue is already in the requested state */
3832                 if (enable == !!(tx_reg & I40E_QTX_ENA_QENA_STAT_MASK))
3833                         continue;
3834
3835                 /* turn on/off the queue */
3836                 if (enable) {
3837                         wr32(hw, I40E_QTX_HEAD(pf_q), 0);
3838                         tx_reg |= I40E_QTX_ENA_QENA_REQ_MASK;
3839                 } else {
3840                         tx_reg &= ~I40E_QTX_ENA_QENA_REQ_MASK;
3841                 }
3842
3843                 wr32(hw, I40E_QTX_ENA(pf_q), tx_reg);
3844                 /* No waiting for the Tx queue to disable */
3845                 if (!enable && test_bit(__I40E_PORT_TX_SUSPENDED, &pf->state))
3846                         continue;
3847
3848                 /* wait for the change to finish */
3849                 ret = i40e_pf_txq_wait(pf, pf_q, enable);
3850                 if (ret) {
3851                         dev_info(&pf->pdev->dev,
3852                                  "VSI seid %d Tx ring %d %sable timeout\n",
3853                                  vsi->seid, pf_q, (enable ? "en" : "dis"));
3854                         break;
3855                 }
3856         }
3857
3858         if (hw->revision_id == 0)
3859                 mdelay(50);
3860         return ret;
3861 }
3862
3863 /**
3864  * i40e_pf_rxq_wait - Wait for a PF's Rx queue to be enabled or disabled
3865  * @pf: the PF being configured
3866  * @pf_q: the PF queue
3867  * @enable: enable or disable state of the queue
3868  *
3869  * This routine will wait for the given Rx queue of the PF to reach the
3870  * enabled or disabled state.
3871  * Returns -ETIMEDOUT in case of failing to reach the requested state after
3872  * multiple retries; else will return 0 in case of success.
3873  **/
3874 static int i40e_pf_rxq_wait(struct i40e_pf *pf, int pf_q, bool enable)
3875 {
3876         int i;
3877         u32 rx_reg;
3878
3879         for (i = 0; i < I40E_QUEUE_WAIT_RETRY_LIMIT; i++) {
3880                 rx_reg = rd32(&pf->hw, I40E_QRX_ENA(pf_q));
3881                 if (enable == !!(rx_reg & I40E_QRX_ENA_QENA_STAT_MASK))
3882                         break;
3883
3884                 usleep_range(10, 20);
3885         }
3886         if (i >= I40E_QUEUE_WAIT_RETRY_LIMIT)
3887                 return -ETIMEDOUT;
3888
3889         return 0;
3890 }
3891
3892 /**
3893  * i40e_vsi_control_rx - Start or stop a VSI's rings
3894  * @vsi: the VSI being configured
3895  * @enable: start or stop the rings
3896  **/
3897 static int i40e_vsi_control_rx(struct i40e_vsi *vsi, bool enable)
3898 {
3899         struct i40e_pf *pf = vsi->back;
3900         struct i40e_hw *hw = &pf->hw;
3901         int i, j, pf_q, ret = 0;
3902         u32 rx_reg;
3903
3904         pf_q = vsi->base_queue;
3905         for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
3906                 for (j = 0; j < 50; j++) {
3907                         rx_reg = rd32(hw, I40E_QRX_ENA(pf_q));
3908                         if (((rx_reg >> I40E_QRX_ENA_QENA_REQ_SHIFT) & 1) ==
3909                             ((rx_reg >> I40E_QRX_ENA_QENA_STAT_SHIFT) & 1))
3910                                 break;
3911                         usleep_range(1000, 2000);
3912                 }
3913
3914                 /* Skip if the queue is already in the requested state */
3915                 if (enable == !!(rx_reg & I40E_QRX_ENA_QENA_STAT_MASK))
3916                         continue;
3917
3918                 /* turn on/off the queue */
3919                 if (enable)
3920                         rx_reg |= I40E_QRX_ENA_QENA_REQ_MASK;
3921                 else
3922                         rx_reg &= ~I40E_QRX_ENA_QENA_REQ_MASK;
3923                 wr32(hw, I40E_QRX_ENA(pf_q), rx_reg);
3924
3925                 /* wait for the change to finish */
3926                 ret = i40e_pf_rxq_wait(pf, pf_q, enable);
3927                 if (ret) {
3928                         dev_info(&pf->pdev->dev,
3929                                  "VSI seid %d Rx ring %d %sable timeout\n",
3930                                  vsi->seid, pf_q, (enable ? "en" : "dis"));
3931                         break;
3932                 }
3933         }
3934
3935         return ret;
3936 }
3937
3938 /**
3939  * i40e_vsi_control_rings - Start or stop a VSI's rings
3940  * @vsi: the VSI being configured
3941  * @enable: start or stop the rings
3942  **/
3943 int i40e_vsi_control_rings(struct i40e_vsi *vsi, bool request)
3944 {
3945         int ret = 0;
3946
3947         /* do rx first for enable and last for disable */
3948         if (request) {
3949                 ret = i40e_vsi_control_rx(vsi, request);
3950                 if (ret)
3951                         return ret;
3952                 ret = i40e_vsi_control_tx(vsi, request);
3953         } else {
3954                 /* Ignore return value, we need to shutdown whatever we can */
3955                 i40e_vsi_control_tx(vsi, request);
3956                 i40e_vsi_control_rx(vsi, request);
3957         }
3958
3959         return ret;
3960 }
3961
3962 /**
3963  * i40e_vsi_free_irq - Free the irq association with the OS
3964  * @vsi: the VSI being configured
3965  **/
3966 static void i40e_vsi_free_irq(struct i40e_vsi *vsi)
3967 {
3968         struct i40e_pf *pf = vsi->back;
3969         struct i40e_hw *hw = &pf->hw;
3970         int base = vsi->base_vector;
3971         u32 val, qp;
3972         int i;
3973
3974         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3975                 if (!vsi->q_vectors)
3976                         return;
3977
3978                 if (!vsi->irqs_ready)
3979                         return;
3980
3981                 vsi->irqs_ready = false;
3982                 for (i = 0; i < vsi->num_q_vectors; i++) {
3983                         u16 vector = i + base;
3984
3985                         /* free only the irqs that were actually requested */
3986                         if (!vsi->q_vectors[i] ||
3987                             !vsi->q_vectors[i]->num_ringpairs)
3988                                 continue;
3989
3990                         /* clear the affinity_mask in the IRQ descriptor */
3991                         irq_set_affinity_hint(pf->msix_entries[vector].vector,
3992                                               NULL);
3993                         free_irq(pf->msix_entries[vector].vector,
3994                                  vsi->q_vectors[i]);
3995
3996                         /* Tear down the interrupt queue link list
3997                          *
3998                          * We know that they come in pairs and always
3999                          * the Rx first, then the Tx.  To clear the
4000                          * link list, stick the EOL value into the
4001                          * next_q field of the registers.
4002                          */
4003                         val = rd32(hw, I40E_PFINT_LNKLSTN(vector - 1));
4004                         qp = (val & I40E_PFINT_LNKLSTN_FIRSTQ_INDX_MASK)
4005                                 >> I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
4006                         val |= I40E_QUEUE_END_OF_LIST
4007                                 << I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
4008                         wr32(hw, I40E_PFINT_LNKLSTN(vector - 1), val);
4009
4010                         while (qp != I40E_QUEUE_END_OF_LIST) {
4011                                 u32 next;
4012
4013                                 val = rd32(hw, I40E_QINT_RQCTL(qp));
4014
4015                                 val &= ~(I40E_QINT_RQCTL_MSIX_INDX_MASK  |
4016                                          I40E_QINT_RQCTL_MSIX0_INDX_MASK |
4017                                          I40E_QINT_RQCTL_CAUSE_ENA_MASK  |
4018                                          I40E_QINT_RQCTL_INTEVENT_MASK);
4019
4020                                 val |= (I40E_QINT_RQCTL_ITR_INDX_MASK |
4021                                          I40E_QINT_RQCTL_NEXTQ_INDX_MASK);
4022
4023                                 wr32(hw, I40E_QINT_RQCTL(qp), val);
4024
4025                                 val = rd32(hw, I40E_QINT_TQCTL(qp));
4026
4027                                 next = (val & I40E_QINT_TQCTL_NEXTQ_INDX_MASK)
4028                                         >> I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT;
4029
4030                                 val &= ~(I40E_QINT_TQCTL_MSIX_INDX_MASK  |
4031                                          I40E_QINT_TQCTL_MSIX0_INDX_MASK |
4032                                          I40E_QINT_TQCTL_CAUSE_ENA_MASK  |
4033                                          I40E_QINT_TQCTL_INTEVENT_MASK);
4034
4035                                 val |= (I40E_QINT_TQCTL_ITR_INDX_MASK |
4036                                          I40E_QINT_TQCTL_NEXTQ_INDX_MASK);
4037
4038                                 wr32(hw, I40E_QINT_TQCTL(qp), val);
4039                                 qp = next;
4040                         }
4041                 }
4042         } else {
4043                 free_irq(pf->pdev->irq, pf);
4044
4045                 val = rd32(hw, I40E_PFINT_LNKLST0);
4046                 qp = (val & I40E_PFINT_LNKLSTN_FIRSTQ_INDX_MASK)
4047                         >> I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
4048                 val |= I40E_QUEUE_END_OF_LIST
4049                         << I40E_PFINT_LNKLST0_FIRSTQ_INDX_SHIFT;
4050                 wr32(hw, I40E_PFINT_LNKLST0, val);
4051
4052                 val = rd32(hw, I40E_QINT_RQCTL(qp));
4053                 val &= ~(I40E_QINT_RQCTL_MSIX_INDX_MASK  |
4054                          I40E_QINT_RQCTL_MSIX0_INDX_MASK |
4055                          I40E_QINT_RQCTL_CAUSE_ENA_MASK  |
4056                          I40E_QINT_RQCTL_INTEVENT_MASK);
4057
4058                 val |= (I40E_QINT_RQCTL_ITR_INDX_MASK |
4059                         I40E_QINT_RQCTL_NEXTQ_INDX_MASK);
4060
4061                 wr32(hw, I40E_QINT_RQCTL(qp), val);
4062
4063                 val = rd32(hw, I40E_QINT_TQCTL(qp));
4064
4065                 val &= ~(I40E_QINT_TQCTL_MSIX_INDX_MASK  |
4066                          I40E_QINT_TQCTL_MSIX0_INDX_MASK |
4067                          I40E_QINT_TQCTL_CAUSE_ENA_MASK  |
4068                          I40E_QINT_TQCTL_INTEVENT_MASK);
4069
4070                 val |= (I40E_QINT_TQCTL_ITR_INDX_MASK |
4071                         I40E_QINT_TQCTL_NEXTQ_INDX_MASK);
4072
4073                 wr32(hw, I40E_QINT_TQCTL(qp), val);
4074         }
4075 }
4076
4077 /**
4078  * i40e_free_q_vector - Free memory allocated for specific interrupt vector
4079  * @vsi: the VSI being configured
4080  * @v_idx: Index of vector to be freed
4081  *
4082  * This function frees the memory allocated to the q_vector.  In addition if
4083  * NAPI is enabled it will delete any references to the NAPI struct prior
4084  * to freeing the q_vector.
4085  **/
4086 static void i40e_free_q_vector(struct i40e_vsi *vsi, int v_idx)
4087 {
4088         struct i40e_q_vector *q_vector = vsi->q_vectors[v_idx];
4089         struct i40e_ring *ring;
4090
4091         if (!q_vector)
4092                 return;
4093
4094         /* disassociate q_vector from rings */
4095         i40e_for_each_ring(ring, q_vector->tx)
4096                 ring->q_vector = NULL;
4097
4098         i40e_for_each_ring(ring, q_vector->rx)
4099                 ring->q_vector = NULL;
4100
4101         /* only VSI w/ an associated netdev is set up w/ NAPI */
4102         if (vsi->netdev)
4103                 netif_napi_del(&q_vector->napi);
4104
4105         vsi->q_vectors[v_idx] = NULL;
4106
4107         kfree_rcu(q_vector, rcu);
4108 }
4109
4110 /**
4111  * i40e_vsi_free_q_vectors - Free memory allocated for interrupt vectors
4112  * @vsi: the VSI being un-configured
4113  *
4114  * This frees the memory allocated to the q_vectors and
4115  * deletes references to the NAPI struct.
4116  **/
4117 static void i40e_vsi_free_q_vectors(struct i40e_vsi *vsi)
4118 {
4119         int v_idx;
4120
4121         for (v_idx = 0; v_idx < vsi->num_q_vectors; v_idx++)
4122                 i40e_free_q_vector(vsi, v_idx);
4123 }
4124
4125 /**
4126  * i40e_reset_interrupt_capability - Disable interrupt setup in OS
4127  * @pf: board private structure
4128  **/
4129 static void i40e_reset_interrupt_capability(struct i40e_pf *pf)
4130 {
4131         /* If we're in Legacy mode, the interrupt was cleaned in vsi_close */
4132         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
4133                 pci_disable_msix(pf->pdev);
4134                 kfree(pf->msix_entries);
4135                 pf->msix_entries = NULL;
4136                 kfree(pf->irq_pile);
4137                 pf->irq_pile = NULL;
4138         } else if (pf->flags & I40E_FLAG_MSI_ENABLED) {
4139                 pci_disable_msi(pf->pdev);
4140         }
4141         pf->flags &= ~(I40E_FLAG_MSIX_ENABLED | I40E_FLAG_MSI_ENABLED);
4142 }
4143
4144 /**
4145  * i40e_clear_interrupt_scheme - Clear the current interrupt scheme settings
4146  * @pf: board private structure
4147  *
4148  * We go through and clear interrupt specific resources and reset the structure
4149  * to pre-load conditions
4150  **/
4151 static void i40e_clear_interrupt_scheme(struct i40e_pf *pf)
4152 {
4153         int i;
4154
4155         i40e_stop_misc_vector(pf);
4156         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
4157                 synchronize_irq(pf->msix_entries[0].vector);
4158                 free_irq(pf->msix_entries[0].vector, pf);
4159         }
4160
4161         i40e_put_lump(pf->irq_pile, 0, I40E_PILE_VALID_BIT-1);
4162         for (i = 0; i < pf->num_alloc_vsi; i++)
4163                 if (pf->vsi[i])
4164                         i40e_vsi_free_q_vectors(pf->vsi[i]);
4165         i40e_reset_interrupt_capability(pf);
4166 }
4167
4168 /**
4169  * i40e_napi_enable_all - Enable NAPI for all q_vectors in the VSI
4170  * @vsi: the VSI being configured
4171  **/
4172 static void i40e_napi_enable_all(struct i40e_vsi *vsi)
4173 {
4174         int q_idx;
4175
4176         if (!vsi->netdev)
4177                 return;
4178
4179         for (q_idx = 0; q_idx < vsi->num_q_vectors; q_idx++)
4180                 napi_enable(&vsi->q_vectors[q_idx]->napi);
4181 }
4182
4183 /**
4184  * i40e_napi_disable_all - Disable NAPI for all q_vectors in the VSI
4185  * @vsi: the VSI being configured
4186  **/
4187 static void i40e_napi_disable_all(struct i40e_vsi *vsi)
4188 {
4189         int q_idx;
4190
4191         if (!vsi->netdev)
4192                 return;
4193
4194         for (q_idx = 0; q_idx < vsi->num_q_vectors; q_idx++)
4195                 napi_disable(&vsi->q_vectors[q_idx]->napi);
4196 }
4197
4198 /**
4199  * i40e_vsi_close - Shut down a VSI
4200  * @vsi: the vsi to be quelled
4201  **/
4202 static void i40e_vsi_close(struct i40e_vsi *vsi)
4203 {
4204         if (!test_and_set_bit(__I40E_DOWN, &vsi->state))
4205                 i40e_down(vsi);
4206         i40e_vsi_free_irq(vsi);
4207         i40e_vsi_free_tx_resources(vsi);
4208         i40e_vsi_free_rx_resources(vsi);
4209         vsi->current_netdev_flags = 0;
4210 }
4211
4212 /**
4213  * i40e_quiesce_vsi - Pause a given VSI
4214  * @vsi: the VSI being paused
4215  **/
4216 static void i40e_quiesce_vsi(struct i40e_vsi *vsi)
4217 {
4218         if (test_bit(__I40E_DOWN, &vsi->state))
4219                 return;
4220
4221         /* No need to disable FCoE VSI when Tx suspended */
4222         if ((test_bit(__I40E_PORT_TX_SUSPENDED, &vsi->back->state)) &&
4223             vsi->type == I40E_VSI_FCOE) {
4224                 dev_dbg(&vsi->back->pdev->dev,
4225                          "VSI seid %d skipping FCoE VSI disable\n", vsi->seid);
4226                 return;
4227         }
4228
4229         set_bit(__I40E_NEEDS_RESTART, &vsi->state);
4230         if (vsi->netdev && netif_running(vsi->netdev))
4231                 vsi->netdev->netdev_ops->ndo_stop(vsi->netdev);
4232         else
4233                 i40e_vsi_close(vsi);
4234 }
4235
4236 /**
4237  * i40e_unquiesce_vsi - Resume a given VSI
4238  * @vsi: the VSI being resumed
4239  **/
4240 static void i40e_unquiesce_vsi(struct i40e_vsi *vsi)
4241 {
4242         if (!test_bit(__I40E_NEEDS_RESTART, &vsi->state))
4243                 return;
4244
4245         clear_bit(__I40E_NEEDS_RESTART, &vsi->state);
4246         if (vsi->netdev && netif_running(vsi->netdev))
4247                 vsi->netdev->netdev_ops->ndo_open(vsi->netdev);
4248         else
4249                 i40e_vsi_open(vsi);   /* this clears the DOWN bit */
4250 }
4251
4252 /**
4253  * i40e_pf_quiesce_all_vsi - Pause all VSIs on a PF
4254  * @pf: the PF
4255  **/
4256 static void i40e_pf_quiesce_all_vsi(struct i40e_pf *pf)
4257 {
4258         int v;
4259
4260         for (v = 0; v < pf->num_alloc_vsi; v++) {
4261                 if (pf->vsi[v])
4262                         i40e_quiesce_vsi(pf->vsi[v]);
4263         }
4264 }
4265
4266 /**
4267  * i40e_pf_unquiesce_all_vsi - Resume all VSIs on a PF
4268  * @pf: the PF
4269  **/
4270 static void i40e_pf_unquiesce_all_vsi(struct i40e_pf *pf)
4271 {
4272         int v;
4273
4274         for (v = 0; v < pf->num_alloc_vsi; v++) {
4275                 if (pf->vsi[v])
4276                         i40e_unquiesce_vsi(pf->vsi[v]);
4277         }
4278 }
4279
4280 #ifdef CONFIG_I40E_DCB
4281 /**
4282  * i40e_vsi_wait_txq_disabled - Wait for VSI's queues to be disabled
4283  * @vsi: the VSI being configured
4284  *
4285  * This function waits for the given VSI's Tx queues to be disabled.
4286  **/
4287 static int i40e_vsi_wait_txq_disabled(struct i40e_vsi *vsi)
4288 {
4289         struct i40e_pf *pf = vsi->back;
4290         int i, pf_q, ret;
4291
4292         pf_q = vsi->base_queue;
4293         for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
4294                 /* Check and wait for the disable status of the queue */
4295                 ret = i40e_pf_txq_wait(pf, pf_q, false);
4296                 if (ret) {
4297                         dev_info(&pf->pdev->dev,
4298                                  "VSI seid %d Tx ring %d disable timeout\n",
4299                                  vsi->seid, pf_q);
4300                         return ret;
4301                 }
4302         }
4303
4304         return 0;
4305 }
4306
4307 /**
4308  * i40e_pf_wait_txq_disabled - Wait for all queues of PF VSIs to be disabled
4309  * @pf: the PF
4310  *
4311  * This function waits for the Tx queues to be in disabled state for all the
4312  * VSIs that are managed by this PF.
4313  **/
4314 static int i40e_pf_wait_txq_disabled(struct i40e_pf *pf)
4315 {
4316         int v, ret = 0;
4317
4318         for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
4319                 /* No need to wait for FCoE VSI queues */
4320                 if (pf->vsi[v] && pf->vsi[v]->type != I40E_VSI_FCOE) {
4321                         ret = i40e_vsi_wait_txq_disabled(pf->vsi[v]);
4322                         if (ret)
4323                                 break;
4324                 }
4325         }
4326
4327         return ret;
4328 }
4329
4330 #endif
4331
4332 /**
4333  * i40e_detect_recover_hung_queue - Function to detect and recover hung_queue
4334  * @q_idx: TX queue number
4335  * @vsi: Pointer to VSI struct
4336  *
4337  * This function checks specified queue for given VSI. Detects hung condition.
4338  * Sets hung bit since it is two step process. Before next run of service task
4339  * if napi_poll runs, it reset 'hung' bit for respective q_vector. If not,
4340  * hung condition remain unchanged and during subsequent run, this function
4341  * issues SW interrupt to recover from hung condition.
4342  **/
4343 static void i40e_detect_recover_hung_queue(int q_idx, struct i40e_vsi *vsi)
4344 {
4345         struct i40e_ring *tx_ring = NULL;
4346         struct i40e_pf  *pf;
4347         u32 head, val, tx_pending;
4348         int i;
4349
4350         pf = vsi->back;
4351
4352         /* now that we have an index, find the tx_ring struct */
4353         for (i = 0; i < vsi->num_queue_pairs; i++) {
4354                 if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc) {
4355                         if (q_idx == vsi->tx_rings[i]->queue_index) {
4356                                 tx_ring = vsi->tx_rings[i];
4357                                 break;
4358                         }
4359                 }
4360         }
4361
4362         if (!tx_ring)
4363                 return;
4364
4365         /* Read interrupt register */
4366         if (pf->flags & I40E_FLAG_MSIX_ENABLED)
4367                 val = rd32(&pf->hw,
4368                            I40E_PFINT_DYN_CTLN(tx_ring->q_vector->v_idx +
4369                                                tx_ring->vsi->base_vector - 1));
4370         else
4371                 val = rd32(&pf->hw, I40E_PFINT_DYN_CTL0);
4372
4373         /* Bail out if interrupts are disabled because napi_poll
4374          * execution in-progress or will get scheduled soon.
4375          * napi_poll cleans TX and RX queues and updates 'next_to_clean'.
4376          */
4377         if (!(val & I40E_PFINT_DYN_CTLN_INTENA_MASK))
4378                 return;
4379
4380         head = i40e_get_head(tx_ring);
4381
4382         tx_pending = i40e_get_tx_pending(tx_ring);
4383
4384         /* HW is done executing descriptors, updated HEAD write back,
4385          * but SW hasn't processed those descriptors. If interrupt is
4386          * not generated from this point ON, it could result into
4387          * dev_watchdog detecting timeout on those netdev_queue,
4388          * hence proactively trigger SW interrupt.
4389          */
4390         if (tx_pending) {
4391                 /* NAPI Poll didn't run and clear since it was set */
4392                 if (test_and_clear_bit(I40E_Q_VECTOR_HUNG_DETECT,
4393                                        &tx_ring->q_vector->hung_detected)) {
4394                         netdev_info(vsi->netdev, "VSI_seid %d, Hung TX queue %d, tx_pending: %d, NTC:0x%x, HWB: 0x%x, NTU: 0x%x, TAIL: 0x%x\n",
4395                                     vsi->seid, q_idx, tx_pending,
4396                                     tx_ring->next_to_clean, head,
4397                                     tx_ring->next_to_use,
4398                                     readl(tx_ring->tail));
4399                         netdev_info(vsi->netdev, "VSI_seid %d, Issuing force_wb for TX queue %d, Interrupt Reg: 0x%x\n",
4400                                     vsi->seid, q_idx, val);
4401                         i40e_force_wb(vsi, tx_ring->q_vector);
4402                 } else {
4403                         /* First Chance - detected possible hung */
4404                         set_bit(I40E_Q_VECTOR_HUNG_DETECT,
4405                                 &tx_ring->q_vector->hung_detected);
4406                 }
4407         }
4408 }
4409
4410 /**
4411  * i40e_detect_recover_hung - Function to detect and recover hung_queues
4412  * @pf:  pointer to PF struct
4413  *
4414  * LAN VSI has netdev and netdev has TX queues. This function is to check
4415  * each of those TX queues if they are hung, trigger recovery by issuing
4416  * SW interrupt.
4417  **/
4418 static void i40e_detect_recover_hung(struct i40e_pf *pf)
4419 {
4420         struct net_device *netdev;
4421         struct i40e_vsi *vsi;
4422         int i;
4423
4424         /* Only for LAN VSI */
4425         vsi = pf->vsi[pf->lan_vsi];
4426
4427         if (!vsi)
4428                 return;
4429
4430         /* Make sure, VSI state is not DOWN/RECOVERY_PENDING */
4431         if (test_bit(__I40E_DOWN, &vsi->back->state) ||
4432             test_bit(__I40E_RESET_RECOVERY_PENDING, &vsi->back->state))
4433                 return;
4434
4435         /* Make sure type is MAIN VSI */
4436         if (vsi->type != I40E_VSI_MAIN)
4437                 return;
4438
4439         netdev = vsi->netdev;
4440         if (!netdev)
4441                 return;
4442
4443         /* Bail out if netif_carrier is not OK */
4444         if (!netif_carrier_ok(netdev))
4445                 return;
4446
4447         /* Go thru' TX queues for netdev */
4448         for (i = 0; i < netdev->num_tx_queues; i++) {
4449                 struct netdev_queue *q;
4450
4451                 q = netdev_get_tx_queue(netdev, i);
4452                 if (q)
4453                         i40e_detect_recover_hung_queue(i, vsi);
4454         }
4455 }
4456
4457 /**
4458  * i40e_get_iscsi_tc_map - Return TC map for iSCSI APP
4459  * @pf: pointer to PF
4460  *
4461  * Get TC map for ISCSI PF type that will include iSCSI TC
4462  * and LAN TC.
4463  **/
4464 static u8 i40e_get_iscsi_tc_map(struct i40e_pf *pf)
4465 {
4466         struct i40e_dcb_app_priority_table app;
4467         struct i40e_hw *hw = &pf->hw;
4468         u8 enabled_tc = 1; /* TC0 is always enabled */
4469         u8 tc, i;
4470         /* Get the iSCSI APP TLV */
4471         struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
4472
4473         for (i = 0; i < dcbcfg->numapps; i++) {
4474                 app = dcbcfg->app[i];
4475                 if (app.selector == I40E_APP_SEL_TCPIP &&
4476                     app.protocolid == I40E_APP_PROTOID_ISCSI) {
4477                         tc = dcbcfg->etscfg.prioritytable[app.priority];
4478                         enabled_tc |= BIT(tc);
4479                         break;
4480                 }
4481         }
4482
4483         return enabled_tc;
4484 }
4485
4486 /**
4487  * i40e_dcb_get_num_tc -  Get the number of TCs from DCBx config
4488  * @dcbcfg: the corresponding DCBx configuration structure
4489  *
4490  * Return the number of TCs from given DCBx configuration
4491  **/
4492 static u8 i40e_dcb_get_num_tc(struct i40e_dcbx_config *dcbcfg)
4493 {
4494         u8 num_tc = 0;
4495         int i;
4496
4497         /* Scan the ETS Config Priority Table to find
4498          * traffic class enabled for a given priority
4499          * and use the traffic class index to get the
4500          * number of traffic classes enabled
4501          */
4502         for (i = 0; i < I40E_MAX_USER_PRIORITY; i++) {
4503                 if (dcbcfg->etscfg.prioritytable[i] > num_tc)
4504                         num_tc = dcbcfg->etscfg.prioritytable[i];
4505         }
4506
4507         /* Traffic class index starts from zero so
4508          * increment to return the actual count
4509          */
4510         return num_tc + 1;
4511 }
4512
4513 /**
4514  * i40e_dcb_get_enabled_tc - Get enabled traffic classes
4515  * @dcbcfg: the corresponding DCBx configuration structure
4516  *
4517  * Query the current DCB configuration and return the number of
4518  * traffic classes enabled from the given DCBX config
4519  **/
4520 static u8 i40e_dcb_get_enabled_tc(struct i40e_dcbx_config *dcbcfg)
4521 {
4522         u8 num_tc = i40e_dcb_get_num_tc(dcbcfg);
4523         u8 enabled_tc = 1;
4524         u8 i;
4525
4526         for (i = 0; i < num_tc; i++)
4527                 enabled_tc |= BIT(i);
4528
4529         return enabled_tc;
4530 }
4531
4532 /**
4533  * i40e_pf_get_num_tc - Get enabled traffic classes for PF
4534  * @pf: PF being queried
4535  *
4536  * Return number of traffic classes enabled for the given PF
4537  **/
4538 static u8 i40e_pf_get_num_tc(struct i40e_pf *pf)
4539 {
4540         struct i40e_hw *hw = &pf->hw;
4541         u8 i, enabled_tc;
4542         u8 num_tc = 0;
4543         struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
4544
4545         /* If DCB is not enabled then always in single TC */
4546         if (!(pf->flags & I40E_FLAG_DCB_ENABLED))
4547                 return 1;
4548
4549         /* SFP mode will be enabled for all TCs on port */
4550         if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
4551                 return i40e_dcb_get_num_tc(dcbcfg);
4552
4553         /* MFP mode return count of enabled TCs for this PF */
4554         if (pf->hw.func_caps.iscsi)
4555                 enabled_tc =  i40e_get_iscsi_tc_map(pf);
4556         else
4557                 return 1; /* Only TC0 */
4558
4559         /* At least have TC0 */
4560         enabled_tc = (enabled_tc ? enabled_tc : 0x1);
4561         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4562                 if (enabled_tc & BIT(i))
4563                         num_tc++;
4564         }
4565         return num_tc;
4566 }
4567
4568 /**
4569  * i40e_pf_get_default_tc - Get bitmap for first enabled TC
4570  * @pf: PF being queried
4571  *
4572  * Return a bitmap for first enabled traffic class for this PF.
4573  **/
4574 static u8 i40e_pf_get_default_tc(struct i40e_pf *pf)
4575 {
4576         u8 enabled_tc = pf->hw.func_caps.enabled_tcmap;
4577         u8 i = 0;
4578
4579         if (!enabled_tc)
4580                 return 0x1; /* TC0 */
4581
4582         /* Find the first enabled TC */
4583         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4584                 if (enabled_tc & BIT(i))
4585                         break;
4586         }
4587
4588         return BIT(i);
4589 }
4590
4591 /**
4592  * i40e_pf_get_pf_tc_map - Get bitmap for enabled traffic classes
4593  * @pf: PF being queried
4594  *
4595  * Return a bitmap for enabled traffic classes for this PF.
4596  **/
4597 static u8 i40e_pf_get_tc_map(struct i40e_pf *pf)
4598 {
4599         /* If DCB is not enabled for this PF then just return default TC */
4600         if (!(pf->flags & I40E_FLAG_DCB_ENABLED))
4601                 return i40e_pf_get_default_tc(pf);
4602
4603         /* SFP mode we want PF to be enabled for all TCs */
4604         if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
4605                 return i40e_dcb_get_enabled_tc(&pf->hw.local_dcbx_config);
4606
4607         /* MFP enabled and iSCSI PF type */
4608         if (pf->hw.func_caps.iscsi)
4609                 return i40e_get_iscsi_tc_map(pf);
4610         else
4611                 return i40e_pf_get_default_tc(pf);
4612 }
4613
4614 /**
4615  * i40e_vsi_get_bw_info - Query VSI BW Information
4616  * @vsi: the VSI being queried
4617  *
4618  * Returns 0 on success, negative value on failure
4619  **/
4620 static int i40e_vsi_get_bw_info(struct i40e_vsi *vsi)
4621 {
4622         struct i40e_aqc_query_vsi_ets_sla_config_resp bw_ets_config = {0};
4623         struct i40e_aqc_query_vsi_bw_config_resp bw_config = {0};
4624         struct i40e_pf *pf = vsi->back;
4625         struct i40e_hw *hw = &pf->hw;
4626         i40e_status ret;
4627         u32 tc_bw_max;
4628         int i;
4629
4630         /* Get the VSI level BW configuration */
4631         ret = i40e_aq_query_vsi_bw_config(hw, vsi->seid, &bw_config, NULL);
4632         if (ret) {
4633                 dev_info(&pf->pdev->dev,
4634                          "couldn't get PF vsi bw config, err %s aq_err %s\n",
4635                          i40e_stat_str(&pf->hw, ret),
4636                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
4637                 return -EINVAL;
4638         }
4639
4640         /* Get the VSI level BW configuration per TC */
4641         ret = i40e_aq_query_vsi_ets_sla_config(hw, vsi->seid, &bw_ets_config,
4642                                                NULL);
4643         if (ret) {
4644                 dev_info(&pf->pdev->dev,
4645                          "couldn't get PF vsi ets bw config, err %s aq_err %s\n",
4646                          i40e_stat_str(&pf->hw, ret),
4647                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
4648                 return -EINVAL;
4649         }
4650
4651         if (bw_config.tc_valid_bits != bw_ets_config.tc_valid_bits) {
4652                 dev_info(&pf->pdev->dev,
4653                          "Enabled TCs mismatch from querying VSI BW info 0x%08x 0x%08x\n",
4654                          bw_config.tc_valid_bits,
4655                          bw_ets_config.tc_valid_bits);
4656                 /* Still continuing */
4657         }
4658
4659         vsi->bw_limit = le16_to_cpu(bw_config.port_bw_limit);
4660         vsi->bw_max_quanta = bw_config.max_bw;
4661         tc_bw_max = le16_to_cpu(bw_ets_config.tc_bw_max[0]) |
4662                     (le16_to_cpu(bw_ets_config.tc_bw_max[1]) << 16);
4663         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4664                 vsi->bw_ets_share_credits[i] = bw_ets_config.share_credits[i];
4665                 vsi->bw_ets_limit_credits[i] =
4666                                         le16_to_cpu(bw_ets_config.credits[i]);
4667                 /* 3 bits out of 4 for each TC */
4668                 vsi->bw_ets_max_quanta[i] = (u8)((tc_bw_max >> (i*4)) & 0x7);
4669         }
4670
4671         return 0;
4672 }
4673
4674 /**
4675  * i40e_vsi_configure_bw_alloc - Configure VSI BW allocation per TC
4676  * @vsi: the VSI being configured
4677  * @enabled_tc: TC bitmap
4678  * @bw_credits: BW shared credits per TC
4679  *
4680  * Returns 0 on success, negative value on failure
4681  **/
4682 static int i40e_vsi_configure_bw_alloc(struct i40e_vsi *vsi, u8 enabled_tc,
4683                                        u8 *bw_share)
4684 {
4685         struct i40e_aqc_configure_vsi_tc_bw_data bw_data;
4686         i40e_status ret;
4687         int i;
4688
4689         bw_data.tc_valid_bits = enabled_tc;
4690         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
4691                 bw_data.tc_bw_credits[i] = bw_share[i];
4692
4693         ret = i40e_aq_config_vsi_tc_bw(&vsi->back->hw, vsi->seid, &bw_data,
4694                                        NULL);
4695         if (ret) {
4696                 dev_info(&vsi->back->pdev->dev,
4697                          "AQ command Config VSI BW allocation per TC failed = %d\n",
4698                          vsi->back->hw.aq.asq_last_status);
4699                 return -EINVAL;
4700         }
4701
4702         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
4703                 vsi->info.qs_handle[i] = bw_data.qs_handles[i];
4704
4705         return 0;
4706 }
4707
4708 /**
4709  * i40e_vsi_config_netdev_tc - Setup the netdev TC configuration
4710  * @vsi: the VSI being configured
4711  * @enabled_tc: TC map to be enabled
4712  *
4713  **/
4714 static void i40e_vsi_config_netdev_tc(struct i40e_vsi *vsi, u8 enabled_tc)
4715 {
4716         struct net_device *netdev = vsi->netdev;
4717         struct i40e_pf *pf = vsi->back;
4718         struct i40e_hw *hw = &pf->hw;
4719         u8 netdev_tc = 0;
4720         int i;
4721         struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
4722
4723         if (!netdev)
4724                 return;
4725
4726         if (!enabled_tc) {
4727                 netdev_reset_tc(netdev);
4728                 return;
4729         }
4730
4731         /* Set up actual enabled TCs on the VSI */
4732         if (netdev_set_num_tc(netdev, vsi->tc_config.numtc))
4733                 return;
4734
4735         /* set per TC queues for the VSI */
4736         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4737                 /* Only set TC queues for enabled tcs
4738                  *
4739                  * e.g. For a VSI that has TC0 and TC3 enabled the
4740                  * enabled_tc bitmap would be 0x00001001; the driver
4741                  * will set the numtc for netdev as 2 that will be
4742                  * referenced by the netdev layer as TC 0 and 1.
4743                  */
4744                 if (vsi->tc_config.enabled_tc & BIT(i))
4745                         netdev_set_tc_queue(netdev,
4746                                         vsi->tc_config.tc_info[i].netdev_tc,
4747                                         vsi->tc_config.tc_info[i].qcount,
4748                                         vsi->tc_config.tc_info[i].qoffset);
4749         }
4750
4751         /* Assign UP2TC map for the VSI */
4752         for (i = 0; i < I40E_MAX_USER_PRIORITY; i++) {
4753                 /* Get the actual TC# for the UP */
4754                 u8 ets_tc = dcbcfg->etscfg.prioritytable[i];
4755                 /* Get the mapped netdev TC# for the UP */
4756                 netdev_tc =  vsi->tc_config.tc_info[ets_tc].netdev_tc;
4757                 netdev_set_prio_tc_map(netdev, i, netdev_tc);
4758         }
4759 }
4760
4761 /**
4762  * i40e_vsi_update_queue_map - Update our copy of VSi info with new queue map
4763  * @vsi: the VSI being configured
4764  * @ctxt: the ctxt buffer returned from AQ VSI update param command
4765  **/
4766 static void i40e_vsi_update_queue_map(struct i40e_vsi *vsi,
4767                                       struct i40e_vsi_context *ctxt)
4768 {
4769         /* copy just the sections touched not the entire info
4770          * since not all sections are valid as returned by
4771          * update vsi params
4772          */
4773         vsi->info.mapping_flags = ctxt->info.mapping_flags;
4774         memcpy(&vsi->info.queue_mapping,
4775                &ctxt->info.queue_mapping, sizeof(vsi->info.queue_mapping));
4776         memcpy(&vsi->info.tc_mapping, ctxt->info.tc_mapping,
4777                sizeof(vsi->info.tc_mapping));
4778 }
4779
4780 /**
4781  * i40e_vsi_config_tc - Configure VSI Tx Scheduler for given TC map
4782  * @vsi: VSI to be configured
4783  * @enabled_tc: TC bitmap
4784  *
4785  * This configures a particular VSI for TCs that are mapped to the
4786  * given TC bitmap. It uses default bandwidth share for TCs across
4787  * VSIs to configure TC for a particular VSI.
4788  *
4789  * NOTE:
4790  * It is expected that the VSI queues have been quisced before calling
4791  * this function.
4792  **/
4793 static int i40e_vsi_config_tc(struct i40e_vsi *vsi, u8 enabled_tc)
4794 {
4795         u8 bw_share[I40E_MAX_TRAFFIC_CLASS] = {0};
4796         struct i40e_vsi_context ctxt;
4797         int ret = 0;
4798         int i;
4799
4800         /* Check if enabled_tc is same as existing or new TCs */
4801         if (vsi->tc_config.enabled_tc == enabled_tc)
4802                 return ret;
4803
4804         /* Enable ETS TCs with equal BW Share for now across all VSIs */
4805         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4806                 if (enabled_tc & BIT(i))
4807                         bw_share[i] = 1;
4808         }
4809
4810         ret = i40e_vsi_configure_bw_alloc(vsi, enabled_tc, bw_share);
4811         if (ret) {
4812                 dev_info(&vsi->back->pdev->dev,
4813                          "Failed configuring TC map %d for VSI %d\n",
4814                          enabled_tc, vsi->seid);
4815                 goto out;
4816         }
4817
4818         /* Update Queue Pairs Mapping for currently enabled UPs */
4819         ctxt.seid = vsi->seid;
4820         ctxt.pf_num = vsi->back->hw.pf_id;
4821         ctxt.vf_num = 0;
4822         ctxt.uplink_seid = vsi->uplink_seid;
4823         ctxt.info = vsi->info;
4824         i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, false);
4825
4826         /* Update the VSI after updating the VSI queue-mapping information */
4827         ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
4828         if (ret) {
4829                 dev_info(&vsi->back->pdev->dev,
4830                          "Update vsi tc config failed, err %s aq_err %s\n",
4831                          i40e_stat_str(&vsi->back->hw, ret),
4832                          i40e_aq_str(&vsi->back->hw,
4833                                      vsi->back->hw.aq.asq_last_status));
4834                 goto out;
4835         }
4836         /* update the local VSI info with updated queue map */
4837         i40e_vsi_update_queue_map(vsi, &ctxt);
4838         vsi->info.valid_sections = 0;
4839
4840         /* Update current VSI BW information */
4841         ret = i40e_vsi_get_bw_info(vsi);
4842         if (ret) {
4843                 dev_info(&vsi->back->pdev->dev,
4844                          "Failed updating vsi bw info, err %s aq_err %s\n",
4845                          i40e_stat_str(&vsi->back->hw, ret),
4846                          i40e_aq_str(&vsi->back->hw,
4847                                      vsi->back->hw.aq.asq_last_status));
4848                 goto out;
4849         }
4850
4851         /* Update the netdev TC setup */
4852         i40e_vsi_config_netdev_tc(vsi, enabled_tc);
4853 out:
4854         return ret;
4855 }
4856
4857 /**
4858  * i40e_veb_config_tc - Configure TCs for given VEB
4859  * @veb: given VEB
4860  * @enabled_tc: TC bitmap
4861  *
4862  * Configures given TC bitmap for VEB (switching) element
4863  **/
4864 int i40e_veb_config_tc(struct i40e_veb *veb, u8 enabled_tc)
4865 {
4866         struct i40e_aqc_configure_switching_comp_bw_config_data bw_data = {0};
4867         struct i40e_pf *pf = veb->pf;
4868         int ret = 0;
4869         int i;
4870
4871         /* No TCs or already enabled TCs just return */
4872         if (!enabled_tc || veb->enabled_tc == enabled_tc)
4873                 return ret;
4874
4875         bw_data.tc_valid_bits = enabled_tc;
4876         /* bw_data.absolute_credits is not set (relative) */
4877
4878         /* Enable ETS TCs with equal BW Share for now */
4879         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4880                 if (enabled_tc & BIT(i))
4881                         bw_data.tc_bw_share_credits[i] = 1;
4882         }
4883
4884         ret = i40e_aq_config_switch_comp_bw_config(&pf->hw, veb->seid,
4885                                                    &bw_data, NULL);
4886         if (ret) {
4887                 dev_info(&pf->pdev->dev,
4888                          "VEB bw config failed, err %s aq_err %s\n",
4889                          i40e_stat_str(&pf->hw, ret),
4890                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
4891                 goto out;
4892         }
4893
4894         /* Update the BW information */
4895         ret = i40e_veb_get_bw_info(veb);
4896         if (ret) {
4897                 dev_info(&pf->pdev->dev,
4898                          "Failed getting veb bw config, err %s aq_err %s\n",
4899                          i40e_stat_str(&pf->hw, ret),
4900                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
4901         }
4902
4903 out:
4904         return ret;
4905 }
4906
4907 #ifdef CONFIG_I40E_DCB
4908 /**
4909  * i40e_dcb_reconfigure - Reconfigure all VEBs and VSIs
4910  * @pf: PF struct
4911  *
4912  * Reconfigure VEB/VSIs on a given PF; it is assumed that
4913  * the caller would've quiesce all the VSIs before calling
4914  * this function
4915  **/
4916 static void i40e_dcb_reconfigure(struct i40e_pf *pf)
4917 {
4918         u8 tc_map = 0;
4919         int ret;
4920         u8 v;
4921
4922         /* Enable the TCs available on PF to all VEBs */
4923         tc_map = i40e_pf_get_tc_map(pf);
4924         for (v = 0; v < I40E_MAX_VEB; v++) {
4925                 if (!pf->veb[v])
4926                         continue;
4927                 ret = i40e_veb_config_tc(pf->veb[v], tc_map);
4928                 if (ret) {
4929                         dev_info(&pf->pdev->dev,
4930                                  "Failed configuring TC for VEB seid=%d\n",
4931                                  pf->veb[v]->seid);
4932                         /* Will try to configure as many components */
4933                 }
4934         }
4935
4936         /* Update each VSI */
4937         for (v = 0; v < pf->num_alloc_vsi; v++) {
4938                 if (!pf->vsi[v])
4939                         continue;
4940
4941                 /* - Enable all TCs for the LAN VSI
4942 #ifdef I40E_FCOE
4943                  * - For FCoE VSI only enable the TC configured
4944                  *   as per the APP TLV
4945 #endif
4946                  * - For all others keep them at TC0 for now
4947                  */
4948                 if (v == pf->lan_vsi)
4949                         tc_map = i40e_pf_get_tc_map(pf);
4950                 else
4951                         tc_map = i40e_pf_get_default_tc(pf);
4952 #ifdef I40E_FCOE
4953                 if (pf->vsi[v]->type == I40E_VSI_FCOE)
4954                         tc_map = i40e_get_fcoe_tc_map(pf);
4955 #endif /* #ifdef I40E_FCOE */
4956
4957                 ret = i40e_vsi_config_tc(pf->vsi[v], tc_map);
4958                 if (ret) {
4959                         dev_info(&pf->pdev->dev,
4960                                  "Failed configuring TC for VSI seid=%d\n",
4961                                  pf->vsi[v]->seid);
4962                         /* Will try to configure as many components */
4963                 } else {
4964                         /* Re-configure VSI vectors based on updated TC map */
4965                         i40e_vsi_map_rings_to_vectors(pf->vsi[v]);
4966                         if (pf->vsi[v]->netdev)
4967                                 i40e_dcbnl_set_all(pf->vsi[v]);
4968                 }
4969         }
4970 }
4971
4972 /**
4973  * i40e_resume_port_tx - Resume port Tx
4974  * @pf: PF struct
4975  *
4976  * Resume a port's Tx and issue a PF reset in case of failure to
4977  * resume.
4978  **/
4979 static int i40e_resume_port_tx(struct i40e_pf *pf)
4980 {
4981         struct i40e_hw *hw = &pf->hw;
4982         int ret;
4983
4984         ret = i40e_aq_resume_port_tx(hw, NULL);
4985         if (ret) {
4986                 dev_info(&pf->pdev->dev,
4987                          "Resume Port Tx failed, err %s aq_err %s\n",
4988                           i40e_stat_str(&pf->hw, ret),
4989                           i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
4990                 /* Schedule PF reset to recover */
4991                 set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
4992                 i40e_service_event_schedule(pf);
4993         }
4994
4995         return ret;
4996 }
4997
4998 /**
4999  * i40e_init_pf_dcb - Initialize DCB configuration
5000  * @pf: PF being configured
5001  *
5002  * Query the current DCB configuration and cache it
5003  * in the hardware structure
5004  **/
5005 static int i40e_init_pf_dcb(struct i40e_pf *pf)
5006 {
5007         struct i40e_hw *hw = &pf->hw;
5008         int err = 0;
5009
5010         /* Do not enable DCB for SW1 and SW2 images even if the FW is capable */
5011         if (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver < 33)) ||
5012             (pf->hw.aq.fw_maj_ver < 4))
5013                 goto out;
5014
5015         /* Get the initial DCB configuration */
5016         err = i40e_init_dcb(hw);
5017         if (!err) {
5018                 /* Device/Function is not DCBX capable */
5019                 if ((!hw->func_caps.dcb) ||
5020                     (hw->dcbx_status == I40E_DCBX_STATUS_DISABLED)) {
5021                         dev_info(&pf->pdev->dev,
5022                                  "DCBX offload is not supported or is disabled for this PF.\n");
5023
5024                         if (pf->flags & I40E_FLAG_MFP_ENABLED)
5025                                 goto out;
5026
5027                 } else {
5028                         /* When status is not DISABLED then DCBX in FW */
5029                         pf->dcbx_cap = DCB_CAP_DCBX_LLD_MANAGED |
5030                                        DCB_CAP_DCBX_VER_IEEE;
5031
5032                         pf->flags |= I40E_FLAG_DCB_CAPABLE;
5033                         /* Enable DCB tagging only when more than one TC */
5034                         if (i40e_dcb_get_num_tc(&hw->local_dcbx_config) > 1)
5035                                 pf->flags |= I40E_FLAG_DCB_ENABLED;
5036                         dev_dbg(&pf->pdev->dev,
5037                                 "DCBX offload is supported for this PF.\n");
5038                 }
5039         } else {
5040                 dev_info(&pf->pdev->dev,
5041                          "Query for DCB configuration failed, err %s aq_err %s\n",
5042                          i40e_stat_str(&pf->hw, err),
5043                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5044         }
5045
5046 out:
5047         return err;
5048 }
5049 #endif /* CONFIG_I40E_DCB */
5050 #define SPEED_SIZE 14
5051 #define FC_SIZE 8
5052 /**
5053  * i40e_print_link_message - print link up or down
5054  * @vsi: the VSI for which link needs a message
5055  */
5056 void i40e_print_link_message(struct i40e_vsi *vsi, bool isup)
5057 {
5058         char *speed = "Unknown";
5059         char *fc = "Unknown";
5060
5061         if (vsi->current_isup == isup)
5062                 return;
5063         vsi->current_isup = isup;
5064         if (!isup) {
5065                 netdev_info(vsi->netdev, "NIC Link is Down\n");
5066                 return;
5067         }
5068
5069         /* Warn user if link speed on NPAR enabled partition is not at
5070          * least 10GB
5071          */
5072         if (vsi->back->hw.func_caps.npar_enable &&
5073             (vsi->back->hw.phy.link_info.link_speed == I40E_LINK_SPEED_1GB ||
5074              vsi->back->hw.phy.link_info.link_speed == I40E_LINK_SPEED_100MB))
5075                 netdev_warn(vsi->netdev,
5076                             "The partition detected link speed that is less than 10Gbps\n");
5077
5078         switch (vsi->back->hw.phy.link_info.link_speed) {
5079         case I40E_LINK_SPEED_40GB:
5080                 speed = "40 G";
5081                 break;
5082         case I40E_LINK_SPEED_20GB:
5083                 speed = "20 G";
5084                 break;
5085         case I40E_LINK_SPEED_10GB:
5086                 speed = "10 G";
5087                 break;
5088         case I40E_LINK_SPEED_1GB:
5089                 speed = "1000 M";
5090                 break;
5091         case I40E_LINK_SPEED_100MB:
5092                 speed = "100 M";
5093                 break;
5094         default:
5095                 break;
5096         }
5097
5098         switch (vsi->back->hw.fc.current_mode) {
5099         case I40E_FC_FULL:
5100                 fc = "RX/TX";
5101                 break;
5102         case I40E_FC_TX_PAUSE:
5103                 fc = "TX";
5104                 break;
5105         case I40E_FC_RX_PAUSE:
5106                 fc = "RX";
5107                 break;
5108         default:
5109                 fc = "None";
5110                 break;
5111         }
5112
5113         netdev_info(vsi->netdev, "NIC Link is Up %sbps Full Duplex, Flow Control: %s\n",
5114                     speed, fc);
5115 }
5116
5117 /**
5118  * i40e_up_complete - Finish the last steps of bringing up a connection
5119  * @vsi: the VSI being configured
5120  **/
5121 static int i40e_up_complete(struct i40e_vsi *vsi)
5122 {
5123         struct i40e_pf *pf = vsi->back;
5124         int err;
5125
5126         if (pf->flags & I40E_FLAG_MSIX_ENABLED)
5127                 i40e_vsi_configure_msix(vsi);
5128         else
5129                 i40e_configure_msi_and_legacy(vsi);
5130
5131         /* start rings */
5132         err = i40e_vsi_control_rings(vsi, true);
5133         if (err)
5134                 return err;
5135
5136         clear_bit(__I40E_DOWN, &vsi->state);
5137         i40e_napi_enable_all(vsi);
5138         i40e_vsi_enable_irq(vsi);
5139
5140         if ((pf->hw.phy.link_info.link_info & I40E_AQ_LINK_UP) &&
5141             (vsi->netdev)) {
5142                 i40e_print_link_message(vsi, true);
5143                 netif_tx_start_all_queues(vsi->netdev);
5144                 netif_carrier_on(vsi->netdev);
5145         } else if (vsi->netdev) {
5146                 i40e_print_link_message(vsi, false);
5147                 /* need to check for qualified module here*/
5148                 if ((pf->hw.phy.link_info.link_info &
5149                         I40E_AQ_MEDIA_AVAILABLE) &&
5150                     (!(pf->hw.phy.link_info.an_info &
5151                         I40E_AQ_QUALIFIED_MODULE)))
5152                         netdev_err(vsi->netdev,
5153                                    "the driver failed to link because an unqualified module was detected.");
5154         }
5155
5156         /* replay FDIR SB filters */
5157         if (vsi->type == I40E_VSI_FDIR) {
5158                 /* reset fd counters */
5159                 pf->fd_add_err = pf->fd_atr_cnt = 0;
5160                 if (pf->fd_tcp_rule > 0) {
5161                         pf->flags &= ~I40E_FLAG_FD_ATR_ENABLED;
5162                         if (I40E_DEBUG_FD & pf->hw.debug_mask)
5163                                 dev_info(&pf->pdev->dev, "Forcing ATR off, sideband rules for TCP/IPv4 exist\n");
5164                         pf->fd_tcp_rule = 0;
5165                 }
5166                 i40e_fdir_filter_restore(vsi);
5167         }
5168         i40e_service_event_schedule(pf);
5169
5170         return 0;
5171 }
5172
5173 /**
5174  * i40e_vsi_reinit_locked - Reset the VSI
5175  * @vsi: the VSI being configured
5176  *
5177  * Rebuild the ring structs after some configuration
5178  * has changed, e.g. MTU size.
5179  **/
5180 static void i40e_vsi_reinit_locked(struct i40e_vsi *vsi)
5181 {
5182         struct i40e_pf *pf = vsi->back;
5183
5184         WARN_ON(in_interrupt());
5185         while (test_and_set_bit(__I40E_CONFIG_BUSY, &pf->state))
5186                 usleep_range(1000, 2000);
5187         i40e_down(vsi);
5188
5189         /* Give a VF some time to respond to the reset.  The
5190          * two second wait is based upon the watchdog cycle in
5191          * the VF driver.
5192          */
5193         if (vsi->type == I40E_VSI_SRIOV)
5194                 msleep(2000);
5195         i40e_up(vsi);
5196         clear_bit(__I40E_CONFIG_BUSY, &pf->state);
5197 }
5198
5199 /**
5200  * i40e_up - Bring the connection back up after being down
5201  * @vsi: the VSI being configured
5202  **/
5203 int i40e_up(struct i40e_vsi *vsi)
5204 {
5205         int err;
5206
5207         err = i40e_vsi_configure(vsi);
5208         if (!err)
5209                 err = i40e_up_complete(vsi);
5210
5211         return err;
5212 }
5213
5214 /**
5215  * i40e_down - Shutdown the connection processing
5216  * @vsi: the VSI being stopped
5217  **/
5218 void i40e_down(struct i40e_vsi *vsi)
5219 {
5220         int i;
5221
5222         /* It is assumed that the caller of this function
5223          * sets the vsi->state __I40E_DOWN bit.
5224          */
5225         if (vsi->netdev) {
5226                 netif_carrier_off(vsi->netdev);
5227                 netif_tx_disable(vsi->netdev);
5228         }
5229         i40e_vsi_disable_irq(vsi);
5230         i40e_vsi_control_rings(vsi, false);
5231         i40e_napi_disable_all(vsi);
5232
5233         for (i = 0; i < vsi->num_queue_pairs; i++) {
5234                 i40e_clean_tx_ring(vsi->tx_rings[i]);
5235                 i40e_clean_rx_ring(vsi->rx_rings[i]);
5236         }
5237 }
5238
5239 /**
5240  * i40e_setup_tc - configure multiple traffic classes
5241  * @netdev: net device to configure
5242  * @tc: number of traffic classes to enable
5243  **/
5244 #ifdef I40E_FCOE
5245 int i40e_setup_tc(struct net_device *netdev, u8 tc)
5246 #else
5247 static int i40e_setup_tc(struct net_device *netdev, u8 tc)
5248 #endif
5249 {
5250         struct i40e_netdev_priv *np = netdev_priv(netdev);
5251         struct i40e_vsi *vsi = np->vsi;
5252         struct i40e_pf *pf = vsi->back;
5253         u8 enabled_tc = 0;
5254         int ret = -EINVAL;
5255         int i;
5256
5257         /* Check if DCB enabled to continue */
5258         if (!(pf->flags & I40E_FLAG_DCB_ENABLED)) {
5259                 netdev_info(netdev, "DCB is not enabled for adapter\n");
5260                 goto exit;
5261         }
5262
5263         /* Check if MFP enabled */
5264         if (pf->flags & I40E_FLAG_MFP_ENABLED) {
5265                 netdev_info(netdev, "Configuring TC not supported in MFP mode\n");
5266                 goto exit;
5267         }
5268
5269         /* Check whether tc count is within enabled limit */
5270         if (tc > i40e_pf_get_num_tc(pf)) {
5271                 netdev_info(netdev, "TC count greater than enabled on link for adapter\n");
5272                 goto exit;
5273         }
5274
5275         /* Generate TC map for number of tc requested */
5276         for (i = 0; i < tc; i++)
5277                 enabled_tc |= BIT(i);
5278
5279         /* Requesting same TC configuration as already enabled */
5280         if (enabled_tc == vsi->tc_config.enabled_tc)
5281                 return 0;
5282
5283         /* Quiesce VSI queues */
5284         i40e_quiesce_vsi(vsi);
5285
5286         /* Configure VSI for enabled TCs */
5287         ret = i40e_vsi_config_tc(vsi, enabled_tc);
5288         if (ret) {
5289                 netdev_info(netdev, "Failed configuring TC for VSI seid=%d\n",
5290                             vsi->seid);
5291                 goto exit;
5292         }
5293
5294         /* Unquiesce VSI */
5295         i40e_unquiesce_vsi(vsi);
5296
5297 exit:
5298         return ret;
5299 }
5300
5301 /**
5302  * i40e_open - Called when a network interface is made active
5303  * @netdev: network interface device structure
5304  *
5305  * The open entry point is called when a network interface is made
5306  * active by the system (IFF_UP).  At this point all resources needed
5307  * for transmit and receive operations are allocated, the interrupt
5308  * handler is registered with the OS, the netdev watchdog subtask is
5309  * enabled, and the stack is notified that the interface is ready.
5310  *
5311  * Returns 0 on success, negative value on failure
5312  **/
5313 int i40e_open(struct net_device *netdev)
5314 {
5315         struct i40e_netdev_priv *np = netdev_priv(netdev);
5316         struct i40e_vsi *vsi = np->vsi;
5317         struct i40e_pf *pf = vsi->back;
5318         int err;
5319
5320         /* disallow open during test or if eeprom is broken */
5321         if (test_bit(__I40E_TESTING, &pf->state) ||
5322             test_bit(__I40E_BAD_EEPROM, &pf->state))
5323                 return -EBUSY;
5324
5325         netif_carrier_off(netdev);
5326
5327         err = i40e_vsi_open(vsi);
5328         if (err)
5329                 return err;
5330
5331         /* configure global TSO hardware offload settings */
5332         wr32(&pf->hw, I40E_GLLAN_TSOMSK_F, be32_to_cpu(TCP_FLAG_PSH |
5333                                                        TCP_FLAG_FIN) >> 16);
5334         wr32(&pf->hw, I40E_GLLAN_TSOMSK_M, be32_to_cpu(TCP_FLAG_PSH |
5335                                                        TCP_FLAG_FIN |
5336                                                        TCP_FLAG_CWR) >> 16);
5337         wr32(&pf->hw, I40E_GLLAN_TSOMSK_L, be32_to_cpu(TCP_FLAG_CWR) >> 16);
5338
5339 #ifdef CONFIG_I40E_VXLAN
5340         vxlan_get_rx_port(netdev);
5341 #endif
5342 #ifdef CONFIG_I40E_GENEVE
5343         geneve_get_rx_port(netdev);
5344 #endif
5345
5346         return 0;
5347 }
5348
5349 /**
5350  * i40e_vsi_open -
5351  * @vsi: the VSI to open
5352  *
5353  * Finish initialization of the VSI.
5354  *
5355  * Returns 0 on success, negative value on failure
5356  **/
5357 int i40e_vsi_open(struct i40e_vsi *vsi)
5358 {
5359         struct i40e_pf *pf = vsi->back;
5360         char int_name[I40E_INT_NAME_STR_LEN];
5361         int err;
5362
5363         /* allocate descriptors */
5364         err = i40e_vsi_setup_tx_resources(vsi);
5365         if (err)
5366                 goto err_setup_tx;
5367         err = i40e_vsi_setup_rx_resources(vsi);
5368         if (err)
5369                 goto err_setup_rx;
5370
5371         err = i40e_vsi_configure(vsi);
5372         if (err)
5373                 goto err_setup_rx;
5374
5375         if (vsi->netdev) {
5376                 snprintf(int_name, sizeof(int_name) - 1, "%s-%s",
5377                          dev_driver_string(&pf->pdev->dev), vsi->netdev->name);
5378                 err = i40e_vsi_request_irq(vsi, int_name);
5379                 if (err)
5380                         goto err_setup_rx;
5381
5382                 /* Notify the stack of the actual queue counts. */
5383                 err = netif_set_real_num_tx_queues(vsi->netdev,
5384                                                    vsi->num_queue_pairs);
5385                 if (err)
5386                         goto err_set_queues;
5387
5388                 err = netif_set_real_num_rx_queues(vsi->netdev,
5389                                                    vsi->num_queue_pairs);
5390                 if (err)
5391                         goto err_set_queues;
5392
5393         } else if (vsi->type == I40E_VSI_FDIR) {
5394                 snprintf(int_name, sizeof(int_name) - 1, "%s-%s:fdir",
5395                          dev_driver_string(&pf->pdev->dev),
5396                          dev_name(&pf->pdev->dev));
5397                 err = i40e_vsi_request_irq(vsi, int_name);
5398
5399         } else {
5400                 err = -EINVAL;
5401                 goto err_setup_rx;
5402         }
5403
5404         err = i40e_up_complete(vsi);
5405         if (err)
5406                 goto err_up_complete;
5407
5408         return 0;
5409
5410 err_up_complete:
5411         i40e_down(vsi);
5412 err_set_queues:
5413         i40e_vsi_free_irq(vsi);
5414 err_setup_rx:
5415         i40e_vsi_free_rx_resources(vsi);
5416 err_setup_tx:
5417         i40e_vsi_free_tx_resources(vsi);
5418         if (vsi == pf->vsi[pf->lan_vsi])
5419                 i40e_do_reset(pf, BIT_ULL(__I40E_PF_RESET_REQUESTED));
5420
5421         return err;
5422 }
5423
5424 /**
5425  * i40e_fdir_filter_exit - Cleans up the Flow Director accounting
5426  * @pf: Pointer to PF
5427  *
5428  * This function destroys the hlist where all the Flow Director
5429  * filters were saved.
5430  **/
5431 static void i40e_fdir_filter_exit(struct i40e_pf *pf)
5432 {
5433         struct i40e_fdir_filter *filter;
5434         struct hlist_node *node2;
5435
5436         hlist_for_each_entry_safe(filter, node2,
5437                                   &pf->fdir_filter_list, fdir_node) {
5438                 hlist_del(&filter->fdir_node);
5439                 kfree(filter);
5440         }
5441         pf->fdir_pf_active_filters = 0;
5442 }
5443
5444 /**
5445  * i40e_close - Disables a network interface
5446  * @netdev: network interface device structure
5447  *
5448  * The close entry point is called when an interface is de-activated
5449  * by the OS.  The hardware is still under the driver's control, but
5450  * this netdev interface is disabled.
5451  *
5452  * Returns 0, this is not allowed to fail
5453  **/
5454 #ifdef I40E_FCOE
5455 int i40e_close(struct net_device *netdev)
5456 #else
5457 static int i40e_close(struct net_device *netdev)
5458 #endif
5459 {
5460         struct i40e_netdev_priv *np = netdev_priv(netdev);
5461         struct i40e_vsi *vsi = np->vsi;
5462
5463         i40e_vsi_close(vsi);
5464
5465         return 0;
5466 }
5467
5468 /**
5469  * i40e_do_reset - Start a PF or Core Reset sequence
5470  * @pf: board private structure
5471  * @reset_flags: which reset is requested
5472  *
5473  * The essential difference in resets is that the PF Reset
5474  * doesn't clear the packet buffers, doesn't reset the PE
5475  * firmware, and doesn't bother the other PFs on the chip.
5476  **/
5477 void i40e_do_reset(struct i40e_pf *pf, u32 reset_flags)
5478 {
5479         u32 val;
5480
5481         WARN_ON(in_interrupt());
5482
5483         if (i40e_check_asq_alive(&pf->hw))
5484                 i40e_vc_notify_reset(pf);
5485
5486         /* do the biggest reset indicated */
5487         if (reset_flags & BIT_ULL(__I40E_GLOBAL_RESET_REQUESTED)) {
5488
5489                 /* Request a Global Reset
5490                  *
5491                  * This will start the chip's countdown to the actual full
5492                  * chip reset event, and a warning interrupt to be sent
5493                  * to all PFs, including the requestor.  Our handler
5494                  * for the warning interrupt will deal with the shutdown
5495                  * and recovery of the switch setup.
5496                  */
5497                 dev_dbg(&pf->pdev->dev, "GlobalR requested\n");
5498                 val = rd32(&pf->hw, I40E_GLGEN_RTRIG);
5499                 val |= I40E_GLGEN_RTRIG_GLOBR_MASK;
5500                 wr32(&pf->hw, I40E_GLGEN_RTRIG, val);
5501
5502         } else if (reset_flags & BIT_ULL(__I40E_CORE_RESET_REQUESTED)) {
5503
5504                 /* Request a Core Reset
5505                  *
5506                  * Same as Global Reset, except does *not* include the MAC/PHY
5507                  */
5508                 dev_dbg(&pf->pdev->dev, "CoreR requested\n");
5509                 val = rd32(&pf->hw, I40E_GLGEN_RTRIG);
5510                 val |= I40E_GLGEN_RTRIG_CORER_MASK;
5511                 wr32(&pf->hw, I40E_GLGEN_RTRIG, val);
5512                 i40e_flush(&pf->hw);
5513
5514         } else if (reset_flags & BIT_ULL(__I40E_PF_RESET_REQUESTED)) {
5515
5516                 /* Request a PF Reset
5517                  *
5518                  * Resets only the PF-specific registers
5519                  *
5520                  * This goes directly to the tear-down and rebuild of
5521                  * the switch, since we need to do all the recovery as
5522                  * for the Core Reset.
5523                  */
5524                 dev_dbg(&pf->pdev->dev, "PFR requested\n");
5525                 i40e_handle_reset_warning(pf);
5526
5527         } else if (reset_flags & BIT_ULL(__I40E_REINIT_REQUESTED)) {
5528                 int v;
5529
5530                 /* Find the VSI(s) that requested a re-init */
5531                 dev_info(&pf->pdev->dev,
5532                          "VSI reinit requested\n");
5533                 for (v = 0; v < pf->num_alloc_vsi; v++) {
5534                         struct i40e_vsi *vsi = pf->vsi[v];
5535
5536                         if (vsi != NULL &&
5537                             test_bit(__I40E_REINIT_REQUESTED, &vsi->state)) {
5538                                 i40e_vsi_reinit_locked(pf->vsi[v]);
5539                                 clear_bit(__I40E_REINIT_REQUESTED, &vsi->state);
5540                         }
5541                 }
5542         } else if (reset_flags & BIT_ULL(__I40E_DOWN_REQUESTED)) {
5543                 int v;
5544
5545                 /* Find the VSI(s) that needs to be brought down */
5546                 dev_info(&pf->pdev->dev, "VSI down requested\n");
5547                 for (v = 0; v < pf->num_alloc_vsi; v++) {
5548                         struct i40e_vsi *vsi = pf->vsi[v];
5549
5550                         if (vsi != NULL &&
5551                             test_bit(__I40E_DOWN_REQUESTED, &vsi->state)) {
5552                                 set_bit(__I40E_DOWN, &vsi->state);
5553                                 i40e_down(vsi);
5554                                 clear_bit(__I40E_DOWN_REQUESTED, &vsi->state);
5555                         }
5556                 }
5557         } else {
5558                 dev_info(&pf->pdev->dev,
5559                          "bad reset request 0x%08x\n", reset_flags);
5560         }
5561 }
5562
5563 #ifdef CONFIG_I40E_DCB
5564 /**
5565  * i40e_dcb_need_reconfig - Check if DCB needs reconfig
5566  * @pf: board private structure
5567  * @old_cfg: current DCB config
5568  * @new_cfg: new DCB config
5569  **/
5570 bool i40e_dcb_need_reconfig(struct i40e_pf *pf,
5571                             struct i40e_dcbx_config *old_cfg,
5572                             struct i40e_dcbx_config *new_cfg)
5573 {
5574         bool need_reconfig = false;
5575
5576         /* Check if ETS configuration has changed */
5577         if (memcmp(&new_cfg->etscfg,
5578                    &old_cfg->etscfg,
5579                    sizeof(new_cfg->etscfg))) {
5580                 /* If Priority Table has changed reconfig is needed */
5581                 if (memcmp(&new_cfg->etscfg.prioritytable,
5582                            &old_cfg->etscfg.prioritytable,
5583                            sizeof(new_cfg->etscfg.prioritytable))) {
5584                         need_reconfig = true;
5585                         dev_dbg(&pf->pdev->dev, "ETS UP2TC changed.\n");
5586                 }
5587
5588                 if (memcmp(&new_cfg->etscfg.tcbwtable,
5589                            &old_cfg->etscfg.tcbwtable,
5590                            sizeof(new_cfg->etscfg.tcbwtable)))
5591                         dev_dbg(&pf->pdev->dev, "ETS TC BW Table changed.\n");
5592
5593                 if (memcmp(&new_cfg->etscfg.tsatable,
5594                            &old_cfg->etscfg.tsatable,
5595                            sizeof(new_cfg->etscfg.tsatable)))
5596                         dev_dbg(&pf->pdev->dev, "ETS TSA Table changed.\n");
5597         }
5598
5599         /* Check if PFC configuration has changed */
5600         if (memcmp(&new_cfg->pfc,
5601                    &old_cfg->pfc,
5602                    sizeof(new_cfg->pfc))) {
5603                 need_reconfig = true;
5604                 dev_dbg(&pf->pdev->dev, "PFC config change detected.\n");
5605         }
5606
5607         /* Check if APP Table has changed */
5608         if (memcmp(&new_cfg->app,
5609                    &old_cfg->app,
5610                    sizeof(new_cfg->app))) {
5611                 need_reconfig = true;
5612                 dev_dbg(&pf->pdev->dev, "APP Table change detected.\n");
5613         }
5614
5615         dev_dbg(&pf->pdev->dev, "dcb need_reconfig=%d\n", need_reconfig);
5616         return need_reconfig;
5617 }
5618
5619 /**
5620  * i40e_handle_lldp_event - Handle LLDP Change MIB event
5621  * @pf: board private structure
5622  * @e: event info posted on ARQ
5623  **/
5624 static int i40e_handle_lldp_event(struct i40e_pf *pf,
5625                                   struct i40e_arq_event_info *e)
5626 {
5627         struct i40e_aqc_lldp_get_mib *mib =
5628                 (struct i40e_aqc_lldp_get_mib *)&e->desc.params.raw;
5629         struct i40e_hw *hw = &pf->hw;
5630         struct i40e_dcbx_config tmp_dcbx_cfg;
5631         bool need_reconfig = false;
5632         int ret = 0;
5633         u8 type;
5634
5635         /* Not DCB capable or capability disabled */
5636         if (!(pf->flags & I40E_FLAG_DCB_CAPABLE))
5637                 return ret;
5638
5639         /* Ignore if event is not for Nearest Bridge */
5640         type = ((mib->type >> I40E_AQ_LLDP_BRIDGE_TYPE_SHIFT)
5641                 & I40E_AQ_LLDP_BRIDGE_TYPE_MASK);
5642         dev_dbg(&pf->pdev->dev, "LLDP event mib bridge type 0x%x\n", type);
5643         if (type != I40E_AQ_LLDP_BRIDGE_TYPE_NEAREST_BRIDGE)
5644                 return ret;
5645
5646         /* Check MIB Type and return if event for Remote MIB update */
5647         type = mib->type & I40E_AQ_LLDP_MIB_TYPE_MASK;
5648         dev_dbg(&pf->pdev->dev,
5649                 "LLDP event mib type %s\n", type ? "remote" : "local");
5650         if (type == I40E_AQ_LLDP_MIB_REMOTE) {
5651                 /* Update the remote cached instance and return */
5652                 ret = i40e_aq_get_dcb_config(hw, I40E_AQ_LLDP_MIB_REMOTE,
5653                                 I40E_AQ_LLDP_BRIDGE_TYPE_NEAREST_BRIDGE,
5654                                 &hw->remote_dcbx_config);
5655                 goto exit;
5656         }
5657
5658         /* Store the old configuration */
5659         tmp_dcbx_cfg = hw->local_dcbx_config;
5660
5661         /* Reset the old DCBx configuration data */
5662         memset(&hw->local_dcbx_config, 0, sizeof(hw->local_dcbx_config));
5663         /* Get updated DCBX data from firmware */
5664         ret = i40e_get_dcb_config(&pf->hw);
5665         if (ret) {
5666                 dev_info(&pf->pdev->dev,
5667                          "Failed querying DCB configuration data from firmware, err %s aq_err %s\n",
5668                          i40e_stat_str(&pf->hw, ret),
5669                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5670                 goto exit;
5671         }
5672
5673         /* No change detected in DCBX configs */
5674         if (!memcmp(&tmp_dcbx_cfg, &hw->local_dcbx_config,
5675                     sizeof(tmp_dcbx_cfg))) {
5676                 dev_dbg(&pf->pdev->dev, "No change detected in DCBX configuration.\n");
5677                 goto exit;
5678         }
5679
5680         need_reconfig = i40e_dcb_need_reconfig(pf, &tmp_dcbx_cfg,
5681                                                &hw->local_dcbx_config);
5682
5683         i40e_dcbnl_flush_apps(pf, &tmp_dcbx_cfg, &hw->local_dcbx_config);
5684
5685         if (!need_reconfig)
5686                 goto exit;
5687
5688         /* Enable DCB tagging only when more than one TC */
5689         if (i40e_dcb_get_num_tc(&hw->local_dcbx_config) > 1)
5690                 pf->flags |= I40E_FLAG_DCB_ENABLED;
5691         else
5692                 pf->flags &= ~I40E_FLAG_DCB_ENABLED;
5693
5694         set_bit(__I40E_PORT_TX_SUSPENDED, &pf->state);
5695         /* Reconfiguration needed quiesce all VSIs */
5696         i40e_pf_quiesce_all_vsi(pf);
5697
5698         /* Changes in configuration update VEB/VSI */
5699         i40e_dcb_reconfigure(pf);
5700
5701         ret = i40e_resume_port_tx(pf);
5702
5703         clear_bit(__I40E_PORT_TX_SUSPENDED, &pf->state);
5704         /* In case of error no point in resuming VSIs */
5705         if (ret)
5706                 goto exit;
5707
5708         /* Wait for the PF's Tx queues to be disabled */
5709         ret = i40e_pf_wait_txq_disabled(pf);
5710         if (ret) {
5711                 /* Schedule PF reset to recover */
5712                 set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
5713                 i40e_service_event_schedule(pf);
5714         } else {
5715                 i40e_pf_unquiesce_all_vsi(pf);
5716         }
5717
5718 exit:
5719         return ret;
5720 }
5721 #endif /* CONFIG_I40E_DCB */
5722
5723 /**
5724  * i40e_do_reset_safe - Protected reset path for userland calls.
5725  * @pf: board private structure
5726  * @reset_flags: which reset is requested
5727  *
5728  **/
5729 void i40e_do_reset_safe(struct i40e_pf *pf, u32 reset_flags)
5730 {
5731         rtnl_lock();
5732         i40e_do_reset(pf, reset_flags);
5733         rtnl_unlock();
5734 }
5735
5736 /**
5737  * i40e_handle_lan_overflow_event - Handler for LAN queue overflow event
5738  * @pf: board private structure
5739  * @e: event info posted on ARQ
5740  *
5741  * Handler for LAN Queue Overflow Event generated by the firmware for PF
5742  * and VF queues
5743  **/
5744 static void i40e_handle_lan_overflow_event(struct i40e_pf *pf,
5745                                            struct i40e_arq_event_info *e)
5746 {
5747         struct i40e_aqc_lan_overflow *data =
5748                 (struct i40e_aqc_lan_overflow *)&e->desc.params.raw;
5749         u32 queue = le32_to_cpu(data->prtdcb_rupto);
5750         u32 qtx_ctl = le32_to_cpu(data->otx_ctl);
5751         struct i40e_hw *hw = &pf->hw;
5752         struct i40e_vf *vf;
5753         u16 vf_id;
5754
5755         dev_dbg(&pf->pdev->dev, "overflow Rx Queue Number = %d QTX_CTL=0x%08x\n",
5756                 queue, qtx_ctl);
5757
5758         /* Queue belongs to VF, find the VF and issue VF reset */
5759         if (((qtx_ctl & I40E_QTX_CTL_PFVF_Q_MASK)
5760             >> I40E_QTX_CTL_PFVF_Q_SHIFT) == I40E_QTX_CTL_VF_QUEUE) {
5761                 vf_id = (u16)((qtx_ctl & I40E_QTX_CTL_VFVM_INDX_MASK)
5762                          >> I40E_QTX_CTL_VFVM_INDX_SHIFT);
5763                 vf_id -= hw->func_caps.vf_base_id;
5764                 vf = &pf->vf[vf_id];
5765                 i40e_vc_notify_vf_reset(vf);
5766                 /* Allow VF to process pending reset notification */
5767                 msleep(20);
5768                 i40e_reset_vf(vf, false);
5769         }
5770 }
5771
5772 /**
5773  * i40e_service_event_complete - Finish up the service event
5774  * @pf: board private structure
5775  **/
5776 static void i40e_service_event_complete(struct i40e_pf *pf)
5777 {
5778         WARN_ON(!test_bit(__I40E_SERVICE_SCHED, &pf->state));
5779
5780         /* flush memory to make sure state is correct before next watchog */
5781         smp_mb__before_atomic();
5782         clear_bit(__I40E_SERVICE_SCHED, &pf->state);
5783 }
5784
5785 /**
5786  * i40e_get_cur_guaranteed_fd_count - Get the consumed guaranteed FD filters
5787  * @pf: board private structure
5788  **/
5789 u32 i40e_get_cur_guaranteed_fd_count(struct i40e_pf *pf)
5790 {
5791         u32 val, fcnt_prog;
5792
5793         val = rd32(&pf->hw, I40E_PFQF_FDSTAT);
5794         fcnt_prog = (val & I40E_PFQF_FDSTAT_GUARANT_CNT_MASK);
5795         return fcnt_prog;
5796 }
5797
5798 /**
5799  * i40e_get_current_fd_count - Get total FD filters programmed for this PF
5800  * @pf: board private structure
5801  **/
5802 u32 i40e_get_current_fd_count(struct i40e_pf *pf)
5803 {
5804         u32 val, fcnt_prog;
5805
5806         val = rd32(&pf->hw, I40E_PFQF_FDSTAT);
5807         fcnt_prog = (val & I40E_PFQF_FDSTAT_GUARANT_CNT_MASK) +
5808                     ((val & I40E_PFQF_FDSTAT_BEST_CNT_MASK) >>
5809                       I40E_PFQF_FDSTAT_BEST_CNT_SHIFT);
5810         return fcnt_prog;
5811 }
5812
5813 /**
5814  * i40e_get_global_fd_count - Get total FD filters programmed on device
5815  * @pf: board private structure
5816  **/
5817 u32 i40e_get_global_fd_count(struct i40e_pf *pf)
5818 {
5819         u32 val, fcnt_prog;
5820
5821         val = rd32(&pf->hw, I40E_GLQF_FDCNT_0);
5822         fcnt_prog = (val & I40E_GLQF_FDCNT_0_GUARANT_CNT_MASK) +
5823                     ((val & I40E_GLQF_FDCNT_0_BESTCNT_MASK) >>
5824                      I40E_GLQF_FDCNT_0_BESTCNT_SHIFT);
5825         return fcnt_prog;
5826 }
5827
5828 /**
5829  * i40e_fdir_check_and_reenable - Function to reenabe FD ATR or SB if disabled
5830  * @pf: board private structure
5831  **/
5832 void i40e_fdir_check_and_reenable(struct i40e_pf *pf)
5833 {
5834         struct i40e_fdir_filter *filter;
5835         u32 fcnt_prog, fcnt_avail;
5836         struct hlist_node *node;
5837
5838         if (test_bit(__I40E_FD_FLUSH_REQUESTED, &pf->state))
5839                 return;
5840
5841         /* Check if, FD SB or ATR was auto disabled and if there is enough room
5842          * to re-enable
5843          */
5844         fcnt_prog = i40e_get_global_fd_count(pf);
5845         fcnt_avail = pf->fdir_pf_filter_count;
5846         if ((fcnt_prog < (fcnt_avail - I40E_FDIR_BUFFER_HEAD_ROOM)) ||
5847             (pf->fd_add_err == 0) ||
5848             (i40e_get_current_atr_cnt(pf) < pf->fd_atr_cnt)) {
5849                 if ((pf->flags & I40E_FLAG_FD_SB_ENABLED) &&
5850                     (pf->auto_disable_flags & I40E_FLAG_FD_SB_ENABLED)) {
5851                         pf->auto_disable_flags &= ~I40E_FLAG_FD_SB_ENABLED;
5852                         if (I40E_DEBUG_FD & pf->hw.debug_mask)
5853                                 dev_info(&pf->pdev->dev, "FD Sideband/ntuple is being enabled since we have space in the table now\n");
5854                 }
5855         }
5856         /* Wait for some more space to be available to turn on ATR */
5857         if (fcnt_prog < (fcnt_avail - I40E_FDIR_BUFFER_HEAD_ROOM * 2)) {
5858                 if ((pf->flags & I40E_FLAG_FD_ATR_ENABLED) &&
5859                     (pf->auto_disable_flags & I40E_FLAG_FD_ATR_ENABLED)) {
5860                         pf->auto_disable_flags &= ~I40E_FLAG_FD_ATR_ENABLED;
5861                         if (I40E_DEBUG_FD & pf->hw.debug_mask)
5862                                 dev_info(&pf->pdev->dev, "ATR is being enabled since we have space in the table now\n");
5863                 }
5864         }
5865
5866         /* if hw had a problem adding a filter, delete it */
5867         if (pf->fd_inv > 0) {
5868                 hlist_for_each_entry_safe(filter, node,
5869                                           &pf->fdir_filter_list, fdir_node) {
5870                         if (filter->fd_id == pf->fd_inv) {
5871                                 hlist_del(&filter->fdir_node);
5872                                 kfree(filter);
5873                                 pf->fdir_pf_active_filters--;
5874                         }
5875                 }
5876         }
5877 }
5878
5879 #define I40E_MIN_FD_FLUSH_INTERVAL 10
5880 #define I40E_MIN_FD_FLUSH_SB_ATR_UNSTABLE 30
5881 /**
5882  * i40e_fdir_flush_and_replay - Function to flush all FD filters and replay SB
5883  * @pf: board private structure
5884  **/
5885 static void i40e_fdir_flush_and_replay(struct i40e_pf *pf)
5886 {
5887         unsigned long min_flush_time;
5888         int flush_wait_retry = 50;
5889         bool disable_atr = false;
5890         int fd_room;
5891         int reg;
5892
5893         if (!(pf->flags & (I40E_FLAG_FD_SB_ENABLED | I40E_FLAG_FD_ATR_ENABLED)))
5894                 return;
5895
5896         if (!time_after(jiffies, pf->fd_flush_timestamp +
5897                                  (I40E_MIN_FD_FLUSH_INTERVAL * HZ)))
5898                 return;
5899
5900         /* If the flush is happening too quick and we have mostly SB rules we
5901          * should not re-enable ATR for some time.
5902          */
5903         min_flush_time = pf->fd_flush_timestamp +
5904                          (I40E_MIN_FD_FLUSH_SB_ATR_UNSTABLE * HZ);
5905         fd_room = pf->fdir_pf_filter_count - pf->fdir_pf_active_filters;
5906
5907         if (!(time_after(jiffies, min_flush_time)) &&
5908             (fd_room < I40E_FDIR_BUFFER_HEAD_ROOM_FOR_ATR)) {
5909                 if (I40E_DEBUG_FD & pf->hw.debug_mask)
5910                         dev_info(&pf->pdev->dev, "ATR disabled, not enough FD filter space.\n");
5911                 disable_atr = true;
5912         }
5913
5914         pf->fd_flush_timestamp = jiffies;
5915         pf->flags &= ~I40E_FLAG_FD_ATR_ENABLED;
5916         /* flush all filters */
5917         wr32(&pf->hw, I40E_PFQF_CTL_1,
5918              I40E_PFQF_CTL_1_CLEARFDTABLE_MASK);
5919         i40e_flush(&pf->hw);
5920         pf->fd_flush_cnt++;
5921         pf->fd_add_err = 0;
5922         do {
5923                 /* Check FD flush status every 5-6msec */
5924                 usleep_range(5000, 6000);
5925                 reg = rd32(&pf->hw, I40E_PFQF_CTL_1);
5926                 if (!(reg & I40E_PFQF_CTL_1_CLEARFDTABLE_MASK))
5927                         break;
5928         } while (flush_wait_retry--);
5929         if (reg & I40E_PFQF_CTL_1_CLEARFDTABLE_MASK) {
5930                 dev_warn(&pf->pdev->dev, "FD table did not flush, needs more time\n");
5931         } else {
5932                 /* replay sideband filters */
5933                 i40e_fdir_filter_restore(pf->vsi[pf->lan_vsi]);
5934                 if (!disable_atr)
5935                         pf->flags |= I40E_FLAG_FD_ATR_ENABLED;
5936                 clear_bit(__I40E_FD_FLUSH_REQUESTED, &pf->state);
5937                 if (I40E_DEBUG_FD & pf->hw.debug_mask)
5938                         dev_info(&pf->pdev->dev, "FD Filter table flushed and FD-SB replayed.\n");
5939         }
5940
5941 }
5942
5943 /**
5944  * i40e_get_current_atr_count - Get the count of total FD ATR filters programmed
5945  * @pf: board private structure
5946  **/
5947 u32 i40e_get_current_atr_cnt(struct i40e_pf *pf)
5948 {
5949         return i40e_get_current_fd_count(pf) - pf->fdir_pf_active_filters;
5950 }
5951
5952 /* We can see up to 256 filter programming desc in transit if the filters are
5953  * being applied really fast; before we see the first
5954  * filter miss error on Rx queue 0. Accumulating enough error messages before
5955  * reacting will make sure we don't cause flush too often.
5956  */
5957 #define I40E_MAX_FD_PROGRAM_ERROR 256
5958
5959 /**
5960  * i40e_fdir_reinit_subtask - Worker thread to reinit FDIR filter table
5961  * @pf: board private structure
5962  **/
5963 static void i40e_fdir_reinit_subtask(struct i40e_pf *pf)
5964 {
5965
5966         /* if interface is down do nothing */
5967         if (test_bit(__I40E_DOWN, &pf->state))
5968                 return;
5969
5970         if (!(pf->flags & (I40E_FLAG_FD_SB_ENABLED | I40E_FLAG_FD_ATR_ENABLED)))
5971                 return;
5972
5973         if (test_bit(__I40E_FD_FLUSH_REQUESTED, &pf->state))
5974                 i40e_fdir_flush_and_replay(pf);
5975
5976         i40e_fdir_check_and_reenable(pf);
5977
5978 }
5979
5980 /**
5981  * i40e_vsi_link_event - notify VSI of a link event
5982  * @vsi: vsi to be notified
5983  * @link_up: link up or down
5984  **/
5985 static void i40e_vsi_link_event(struct i40e_vsi *vsi, bool link_up)
5986 {
5987         if (!vsi || test_bit(__I40E_DOWN, &vsi->state))
5988                 return;
5989
5990         switch (vsi->type) {
5991         case I40E_VSI_MAIN:
5992 #ifdef I40E_FCOE
5993         case I40E_VSI_FCOE:
5994 #endif
5995                 if (!vsi->netdev || !vsi->netdev_registered)
5996                         break;
5997
5998                 if (link_up) {
5999                         netif_carrier_on(vsi->netdev);
6000                         netif_tx_wake_all_queues(vsi->netdev);
6001                 } else {
6002                         netif_carrier_off(vsi->netdev);
6003                         netif_tx_stop_all_queues(vsi->netdev);
6004                 }
6005                 break;
6006
6007         case I40E_VSI_SRIOV:
6008         case I40E_VSI_VMDQ2:
6009         case I40E_VSI_CTRL:
6010         case I40E_VSI_MIRROR:
6011         default:
6012                 /* there is no notification for other VSIs */
6013                 break;
6014         }
6015 }
6016
6017 /**
6018  * i40e_veb_link_event - notify elements on the veb of a link event
6019  * @veb: veb to be notified
6020  * @link_up: link up or down
6021  **/
6022 static void i40e_veb_link_event(struct i40e_veb *veb, bool link_up)
6023 {
6024         struct i40e_pf *pf;
6025         int i;
6026
6027         if (!veb || !veb->pf)
6028                 return;
6029         pf = veb->pf;
6030
6031         /* depth first... */
6032         for (i = 0; i < I40E_MAX_VEB; i++)
6033                 if (pf->veb[i] && (pf->veb[i]->uplink_seid == veb->seid))
6034                         i40e_veb_link_event(pf->veb[i], link_up);
6035
6036         /* ... now the local VSIs */
6037         for (i = 0; i < pf->num_alloc_vsi; i++)
6038                 if (pf->vsi[i] && (pf->vsi[i]->uplink_seid == veb->seid))
6039                         i40e_vsi_link_event(pf->vsi[i], link_up);
6040 }
6041
6042 /**
6043  * i40e_link_event - Update netif_carrier status
6044  * @pf: board private structure
6045  **/
6046 static void i40e_link_event(struct i40e_pf *pf)
6047 {
6048         struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
6049         u8 new_link_speed, old_link_speed;
6050         i40e_status status;
6051         bool new_link, old_link;
6052
6053         /* save off old link status information */
6054         pf->hw.phy.link_info_old = pf->hw.phy.link_info;
6055
6056         /* set this to force the get_link_status call to refresh state */
6057         pf->hw.phy.get_link_info = true;
6058
6059         old_link = (pf->hw.phy.link_info_old.link_info & I40E_AQ_LINK_UP);
6060
6061         status = i40e_get_link_status(&pf->hw, &new_link);
6062         if (status) {
6063                 dev_dbg(&pf->pdev->dev, "couldn't get link state, status: %d\n",
6064                         status);
6065                 return;
6066         }
6067
6068         old_link_speed = pf->hw.phy.link_info_old.link_speed;
6069         new_link_speed = pf->hw.phy.link_info.link_speed;
6070
6071         if (new_link == old_link &&
6072             new_link_speed == old_link_speed &&
6073             (test_bit(__I40E_DOWN, &vsi->state) ||
6074              new_link == netif_carrier_ok(vsi->netdev)))
6075                 return;
6076
6077         if (!test_bit(__I40E_DOWN, &vsi->state))
6078                 i40e_print_link_message(vsi, new_link);
6079
6080         /* Notify the base of the switch tree connected to
6081          * the link.  Floating VEBs are not notified.
6082          */
6083         if (pf->lan_veb != I40E_NO_VEB && pf->veb[pf->lan_veb])
6084                 i40e_veb_link_event(pf->veb[pf->lan_veb], new_link);
6085         else
6086                 i40e_vsi_link_event(vsi, new_link);
6087
6088         if (pf->vf)
6089                 i40e_vc_notify_link_state(pf);
6090
6091         if (pf->flags & I40E_FLAG_PTP)
6092                 i40e_ptp_set_increment(pf);
6093 }
6094
6095 /**
6096  * i40e_watchdog_subtask - periodic checks not using event driven response
6097  * @pf: board private structure
6098  **/
6099 static void i40e_watchdog_subtask(struct i40e_pf *pf)
6100 {
6101         int i;
6102
6103         /* if interface is down do nothing */
6104         if (test_bit(__I40E_DOWN, &pf->state) ||
6105             test_bit(__I40E_CONFIG_BUSY, &pf->state))
6106                 return;
6107
6108         /* make sure we don't do these things too often */
6109         if (time_before(jiffies, (pf->service_timer_previous +
6110                                   pf->service_timer_period)))
6111                 return;
6112         pf->service_timer_previous = jiffies;
6113
6114         if (pf->flags & I40E_FLAG_LINK_POLLING_ENABLED)
6115                 i40e_link_event(pf);
6116
6117         /* Update the stats for active netdevs so the network stack
6118          * can look at updated numbers whenever it cares to
6119          */
6120         for (i = 0; i < pf->num_alloc_vsi; i++)
6121                 if (pf->vsi[i] && pf->vsi[i]->netdev)
6122                         i40e_update_stats(pf->vsi[i]);
6123
6124         if (pf->flags & I40E_FLAG_VEB_STATS_ENABLED) {
6125                 /* Update the stats for the active switching components */
6126                 for (i = 0; i < I40E_MAX_VEB; i++)
6127                         if (pf->veb[i])
6128                                 i40e_update_veb_stats(pf->veb[i]);
6129         }
6130
6131         i40e_ptp_rx_hang(pf->vsi[pf->lan_vsi]);
6132 }
6133
6134 /**
6135  * i40e_reset_subtask - Set up for resetting the device and driver
6136  * @pf: board private structure
6137  **/
6138 static void i40e_reset_subtask(struct i40e_pf *pf)
6139 {
6140         u32 reset_flags = 0;
6141
6142         rtnl_lock();
6143         if (test_bit(__I40E_REINIT_REQUESTED, &pf->state)) {
6144                 reset_flags |= BIT(__I40E_REINIT_REQUESTED);
6145                 clear_bit(__I40E_REINIT_REQUESTED, &pf->state);
6146         }
6147         if (test_bit(__I40E_PF_RESET_REQUESTED, &pf->state)) {
6148                 reset_flags |= BIT(__I40E_PF_RESET_REQUESTED);
6149                 clear_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
6150         }
6151         if (test_bit(__I40E_CORE_RESET_REQUESTED, &pf->state)) {
6152                 reset_flags |= BIT(__I40E_CORE_RESET_REQUESTED);
6153                 clear_bit(__I40E_CORE_RESET_REQUESTED, &pf->state);
6154         }
6155         if (test_bit(__I40E_GLOBAL_RESET_REQUESTED, &pf->state)) {
6156                 reset_flags |= BIT(__I40E_GLOBAL_RESET_REQUESTED);
6157                 clear_bit(__I40E_GLOBAL_RESET_REQUESTED, &pf->state);
6158         }
6159         if (test_bit(__I40E_DOWN_REQUESTED, &pf->state)) {
6160                 reset_flags |= BIT(__I40E_DOWN_REQUESTED);
6161                 clear_bit(__I40E_DOWN_REQUESTED, &pf->state);
6162         }
6163
6164         /* If there's a recovery already waiting, it takes
6165          * precedence before starting a new reset sequence.
6166          */
6167         if (test_bit(__I40E_RESET_INTR_RECEIVED, &pf->state)) {
6168                 i40e_handle_reset_warning(pf);
6169                 goto unlock;
6170         }
6171
6172         /* If we're already down or resetting, just bail */
6173         if (reset_flags &&
6174             !test_bit(__I40E_DOWN, &pf->state) &&
6175             !test_bit(__I40E_CONFIG_BUSY, &pf->state))
6176                 i40e_do_reset(pf, reset_flags);
6177
6178 unlock:
6179         rtnl_unlock();
6180 }
6181
6182 /**
6183  * i40e_handle_link_event - Handle link event
6184  * @pf: board private structure
6185  * @e: event info posted on ARQ
6186  **/
6187 static void i40e_handle_link_event(struct i40e_pf *pf,
6188                                    struct i40e_arq_event_info *e)
6189 {
6190         struct i40e_aqc_get_link_status *status =
6191                 (struct i40e_aqc_get_link_status *)&e->desc.params.raw;
6192
6193         /* Do a new status request to re-enable LSE reporting
6194          * and load new status information into the hw struct
6195          * This completely ignores any state information
6196          * in the ARQ event info, instead choosing to always
6197          * issue the AQ update link status command.
6198          */
6199         i40e_link_event(pf);
6200
6201         /* check for unqualified module, if link is down */
6202         if ((status->link_info & I40E_AQ_MEDIA_AVAILABLE) &&
6203             (!(status->an_info & I40E_AQ_QUALIFIED_MODULE)) &&
6204             (!(status->link_info & I40E_AQ_LINK_UP)))
6205                 dev_err(&pf->pdev->dev,
6206                         "The driver failed to link because an unqualified module was detected.\n");
6207 }
6208
6209 /**
6210  * i40e_clean_adminq_subtask - Clean the AdminQ rings
6211  * @pf: board private structure
6212  **/
6213 static void i40e_clean_adminq_subtask(struct i40e_pf *pf)
6214 {
6215         struct i40e_arq_event_info event;
6216         struct i40e_hw *hw = &pf->hw;
6217         u16 pending, i = 0;
6218         i40e_status ret;
6219         u16 opcode;
6220         u32 oldval;
6221         u32 val;
6222
6223         /* Do not run clean AQ when PF reset fails */
6224         if (test_bit(__I40E_RESET_FAILED, &pf->state))
6225                 return;
6226
6227         /* check for error indications */
6228         val = rd32(&pf->hw, pf->hw.aq.arq.len);
6229         oldval = val;
6230         if (val & I40E_PF_ARQLEN_ARQVFE_MASK) {
6231                 if (hw->debug_mask & I40E_DEBUG_AQ)
6232                         dev_info(&pf->pdev->dev, "ARQ VF Error detected\n");
6233                 val &= ~I40E_PF_ARQLEN_ARQVFE_MASK;
6234         }
6235         if (val & I40E_PF_ARQLEN_ARQOVFL_MASK) {
6236                 if (hw->debug_mask & I40E_DEBUG_AQ)
6237                         dev_info(&pf->pdev->dev, "ARQ Overflow Error detected\n");
6238                 val &= ~I40E_PF_ARQLEN_ARQOVFL_MASK;
6239         }
6240         if (val & I40E_PF_ARQLEN_ARQCRIT_MASK) {
6241                 if (hw->debug_mask & I40E_DEBUG_AQ)
6242                         dev_info(&pf->pdev->dev, "ARQ Critical Error detected\n");
6243                 val &= ~I40E_PF_ARQLEN_ARQCRIT_MASK;
6244         }
6245         if (oldval != val)
6246                 wr32(&pf->hw, pf->hw.aq.arq.len, val);
6247
6248         val = rd32(&pf->hw, pf->hw.aq.asq.len);
6249         oldval = val;
6250         if (val & I40E_PF_ATQLEN_ATQVFE_MASK) {
6251                 if (pf->hw.debug_mask & I40E_DEBUG_AQ)
6252                         dev_info(&pf->pdev->dev, "ASQ VF Error detected\n");
6253                 val &= ~I40E_PF_ATQLEN_ATQVFE_MASK;
6254         }
6255         if (val & I40E_PF_ATQLEN_ATQOVFL_MASK) {
6256                 if (pf->hw.debug_mask & I40E_DEBUG_AQ)
6257                         dev_info(&pf->pdev->dev, "ASQ Overflow Error detected\n");
6258                 val &= ~I40E_PF_ATQLEN_ATQOVFL_MASK;
6259         }
6260         if (val & I40E_PF_ATQLEN_ATQCRIT_MASK) {
6261                 if (pf->hw.debug_mask & I40E_DEBUG_AQ)
6262                         dev_info(&pf->pdev->dev, "ASQ Critical Error detected\n");
6263                 val &= ~I40E_PF_ATQLEN_ATQCRIT_MASK;
6264         }
6265         if (oldval != val)
6266                 wr32(&pf->hw, pf->hw.aq.asq.len, val);
6267
6268         event.buf_len = I40E_MAX_AQ_BUF_SIZE;
6269         event.msg_buf = kzalloc(event.buf_len, GFP_KERNEL);
6270         if (!event.msg_buf)
6271                 return;
6272
6273         do {
6274                 ret = i40e_clean_arq_element(hw, &event, &pending);
6275                 if (ret == I40E_ERR_ADMIN_QUEUE_NO_WORK)
6276                         break;
6277                 else if (ret) {
6278                         dev_info(&pf->pdev->dev, "ARQ event error %d\n", ret);
6279                         break;
6280                 }
6281
6282                 opcode = le16_to_cpu(event.desc.opcode);
6283                 switch (opcode) {
6284
6285                 case i40e_aqc_opc_get_link_status:
6286                         i40e_handle_link_event(pf, &event);
6287                         break;
6288                 case i40e_aqc_opc_send_msg_to_pf:
6289                         ret = i40e_vc_process_vf_msg(pf,
6290                                         le16_to_cpu(event.desc.retval),
6291                                         le32_to_cpu(event.desc.cookie_high),
6292                                         le32_to_cpu(event.desc.cookie_low),
6293                                         event.msg_buf,
6294                                         event.msg_len);
6295                         break;
6296                 case i40e_aqc_opc_lldp_update_mib:
6297                         dev_dbg(&pf->pdev->dev, "ARQ: Update LLDP MIB event received\n");
6298 #ifdef CONFIG_I40E_DCB
6299                         rtnl_lock();
6300                         ret = i40e_handle_lldp_event(pf, &event);
6301                         rtnl_unlock();
6302 #endif /* CONFIG_I40E_DCB */
6303                         break;
6304                 case i40e_aqc_opc_event_lan_overflow:
6305                         dev_dbg(&pf->pdev->dev, "ARQ LAN queue overflow event received\n");
6306                         i40e_handle_lan_overflow_event(pf, &event);
6307                         break;
6308                 case i40e_aqc_opc_send_msg_to_peer:
6309                         dev_info(&pf->pdev->dev, "ARQ: Msg from other pf\n");
6310                         break;
6311                 case i40e_aqc_opc_nvm_erase:
6312                 case i40e_aqc_opc_nvm_update:
6313                 case i40e_aqc_opc_oem_post_update:
6314                         i40e_debug(&pf->hw, I40E_DEBUG_NVM, "ARQ NVM operation completed\n");
6315                         break;
6316                 default:
6317                         dev_info(&pf->pdev->dev,
6318                                  "ARQ Error: Unknown event 0x%04x received\n",
6319                                  opcode);
6320                         break;
6321                 }
6322         } while (pending && (i++ < pf->adminq_work_limit));
6323
6324         clear_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state);
6325         /* re-enable Admin queue interrupt cause */
6326         val = rd32(hw, I40E_PFINT_ICR0_ENA);
6327         val |=  I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
6328         wr32(hw, I40E_PFINT_ICR0_ENA, val);
6329         i40e_flush(hw);
6330
6331         kfree(event.msg_buf);
6332 }
6333
6334 /**
6335  * i40e_verify_eeprom - make sure eeprom is good to use
6336  * @pf: board private structure
6337  **/
6338 static void i40e_verify_eeprom(struct i40e_pf *pf)
6339 {
6340         int err;
6341
6342         err = i40e_diag_eeprom_test(&pf->hw);
6343         if (err) {
6344                 /* retry in case of garbage read */
6345                 err = i40e_diag_eeprom_test(&pf->hw);
6346                 if (err) {
6347                         dev_info(&pf->pdev->dev, "eeprom check failed (%d), Tx/Rx traffic disabled\n",
6348                                  err);
6349                         set_bit(__I40E_BAD_EEPROM, &pf->state);
6350                 }
6351         }
6352
6353         if (!err && test_bit(__I40E_BAD_EEPROM, &pf->state)) {
6354                 dev_info(&pf->pdev->dev, "eeprom check passed, Tx/Rx traffic enabled\n");
6355                 clear_bit(__I40E_BAD_EEPROM, &pf->state);
6356         }
6357 }
6358
6359 /**
6360  * i40e_enable_pf_switch_lb
6361  * @pf: pointer to the PF structure
6362  *
6363  * enable switch loop back or die - no point in a return value
6364  **/
6365 static void i40e_enable_pf_switch_lb(struct i40e_pf *pf)
6366 {
6367         struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
6368         struct i40e_vsi_context ctxt;
6369         int ret;
6370
6371         ctxt.seid = pf->main_vsi_seid;
6372         ctxt.pf_num = pf->hw.pf_id;
6373         ctxt.vf_num = 0;
6374         ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
6375         if (ret) {
6376                 dev_info(&pf->pdev->dev,
6377                          "couldn't get PF vsi config, err %s aq_err %s\n",
6378                          i40e_stat_str(&pf->hw, ret),
6379                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6380                 return;
6381         }
6382         ctxt.flags = I40E_AQ_VSI_TYPE_PF;
6383         ctxt.info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
6384         ctxt.info.switch_id |= cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
6385
6386         ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
6387         if (ret) {
6388                 dev_info(&pf->pdev->dev,
6389                          "update vsi switch failed, err %s aq_err %s\n",
6390                          i40e_stat_str(&pf->hw, ret),
6391                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6392         }
6393 }
6394
6395 /**
6396  * i40e_disable_pf_switch_lb
6397  * @pf: pointer to the PF structure
6398  *
6399  * disable switch loop back or die - no point in a return value
6400  **/
6401 static void i40e_disable_pf_switch_lb(struct i40e_pf *pf)
6402 {
6403         struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
6404         struct i40e_vsi_context ctxt;
6405         int ret;
6406
6407         ctxt.seid = pf->main_vsi_seid;
6408         ctxt.pf_num = pf->hw.pf_id;
6409         ctxt.vf_num = 0;
6410         ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
6411         if (ret) {
6412                 dev_info(&pf->pdev->dev,
6413                          "couldn't get PF vsi config, err %s aq_err %s\n",
6414                          i40e_stat_str(&pf->hw, ret),
6415                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6416                 return;
6417         }
6418         ctxt.flags = I40E_AQ_VSI_TYPE_PF;
6419         ctxt.info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
6420         ctxt.info.switch_id &= ~cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
6421
6422         ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
6423         if (ret) {
6424                 dev_info(&pf->pdev->dev,
6425                          "update vsi switch failed, err %s aq_err %s\n",
6426                          i40e_stat_str(&pf->hw, ret),
6427                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6428         }
6429 }
6430
6431 /**
6432  * i40e_config_bridge_mode - Configure the HW bridge mode
6433  * @veb: pointer to the bridge instance
6434  *
6435  * Configure the loop back mode for the LAN VSI that is downlink to the
6436  * specified HW bridge instance. It is expected this function is called
6437  * when a new HW bridge is instantiated.
6438  **/
6439 static void i40e_config_bridge_mode(struct i40e_veb *veb)
6440 {
6441         struct i40e_pf *pf = veb->pf;
6442
6443         if (pf->hw.debug_mask & I40E_DEBUG_LAN)
6444                 dev_info(&pf->pdev->dev, "enabling bridge mode: %s\n",
6445                          veb->bridge_mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
6446         if (veb->bridge_mode & BRIDGE_MODE_VEPA)
6447                 i40e_disable_pf_switch_lb(pf);
6448         else
6449                 i40e_enable_pf_switch_lb(pf);
6450 }
6451
6452 /**
6453  * i40e_reconstitute_veb - rebuild the VEB and anything connected to it
6454  * @veb: pointer to the VEB instance
6455  *
6456  * This is a recursive function that first builds the attached VSIs then
6457  * recurses in to build the next layer of VEB.  We track the connections
6458  * through our own index numbers because the seid's from the HW could
6459  * change across the reset.
6460  **/
6461 static int i40e_reconstitute_veb(struct i40e_veb *veb)
6462 {
6463         struct i40e_vsi *ctl_vsi = NULL;
6464         struct i40e_pf *pf = veb->pf;
6465         int v, veb_idx;
6466         int ret;
6467
6468         /* build VSI that owns this VEB, temporarily attached to base VEB */
6469         for (v = 0; v < pf->num_alloc_vsi && !ctl_vsi; v++) {
6470                 if (pf->vsi[v] &&
6471                     pf->vsi[v]->veb_idx == veb->idx &&
6472                     pf->vsi[v]->flags & I40E_VSI_FLAG_VEB_OWNER) {
6473                         ctl_vsi = pf->vsi[v];
6474                         break;
6475                 }
6476         }
6477         if (!ctl_vsi) {
6478                 dev_info(&pf->pdev->dev,
6479                          "missing owner VSI for veb_idx %d\n", veb->idx);
6480                 ret = -ENOENT;
6481                 goto end_reconstitute;
6482         }
6483         if (ctl_vsi != pf->vsi[pf->lan_vsi])
6484                 ctl_vsi->uplink_seid = pf->vsi[pf->lan_vsi]->uplink_seid;
6485         ret = i40e_add_vsi(ctl_vsi);
6486         if (ret) {
6487                 dev_info(&pf->pdev->dev,
6488                          "rebuild of veb_idx %d owner VSI failed: %d\n",
6489                          veb->idx, ret);
6490                 goto end_reconstitute;
6491         }
6492         i40e_vsi_reset_stats(ctl_vsi);
6493
6494         /* create the VEB in the switch and move the VSI onto the VEB */
6495         ret = i40e_add_veb(veb, ctl_vsi);
6496         if (ret)
6497                 goto end_reconstitute;
6498
6499         if (pf->flags & I40E_FLAG_VEB_MODE_ENABLED)
6500                 veb->bridge_mode = BRIDGE_MODE_VEB;
6501         else
6502                 veb->bridge_mode = BRIDGE_MODE_VEPA;
6503         i40e_config_bridge_mode(veb);
6504
6505         /* create the remaining VSIs attached to this VEB */
6506         for (v = 0; v < pf->num_alloc_vsi; v++) {
6507                 if (!pf->vsi[v] || pf->vsi[v] == ctl_vsi)
6508                         continue;
6509
6510                 if (pf->vsi[v]->veb_idx == veb->idx) {
6511                         struct i40e_vsi *vsi = pf->vsi[v];
6512
6513                         vsi->uplink_seid = veb->seid;
6514                         ret = i40e_add_vsi(vsi);
6515                         if (ret) {
6516                                 dev_info(&pf->pdev->dev,
6517                                          "rebuild of vsi_idx %d failed: %d\n",
6518                                          v, ret);
6519                                 goto end_reconstitute;
6520                         }
6521                         i40e_vsi_reset_stats(vsi);
6522                 }
6523         }
6524
6525         /* create any VEBs attached to this VEB - RECURSION */
6526         for (veb_idx = 0; veb_idx < I40E_MAX_VEB; veb_idx++) {
6527                 if (pf->veb[veb_idx] && pf->veb[veb_idx]->veb_idx == veb->idx) {
6528                         pf->veb[veb_idx]->uplink_seid = veb->seid;
6529                         ret = i40e_reconstitute_veb(pf->veb[veb_idx]);
6530                         if (ret)
6531                                 break;
6532                 }
6533         }
6534
6535 end_reconstitute:
6536         return ret;
6537 }
6538
6539 /**
6540  * i40e_get_capabilities - get info about the HW
6541  * @pf: the PF struct
6542  **/
6543 static int i40e_get_capabilities(struct i40e_pf *pf)
6544 {
6545         struct i40e_aqc_list_capabilities_element_resp *cap_buf;
6546         u16 data_size;
6547         int buf_len;
6548         int err;
6549
6550         buf_len = 40 * sizeof(struct i40e_aqc_list_capabilities_element_resp);
6551         do {
6552                 cap_buf = kzalloc(buf_len, GFP_KERNEL);
6553                 if (!cap_buf)
6554                         return -ENOMEM;
6555
6556                 /* this loads the data into the hw struct for us */
6557                 err = i40e_aq_discover_capabilities(&pf->hw, cap_buf, buf_len,
6558                                             &data_size,
6559                                             i40e_aqc_opc_list_func_capabilities,
6560                                             NULL);
6561                 /* data loaded, buffer no longer needed */
6562                 kfree(cap_buf);
6563
6564                 if (pf->hw.aq.asq_last_status == I40E_AQ_RC_ENOMEM) {
6565                         /* retry with a larger buffer */
6566                         buf_len = data_size;
6567                 } else if (pf->hw.aq.asq_last_status != I40E_AQ_RC_OK) {
6568                         dev_info(&pf->pdev->dev,
6569                                  "capability discovery failed, err %s aq_err %s\n",
6570                                  i40e_stat_str(&pf->hw, err),
6571                                  i40e_aq_str(&pf->hw,
6572                                              pf->hw.aq.asq_last_status));
6573                         return -ENODEV;
6574                 }
6575         } while (err);
6576
6577         if (pf->hw.debug_mask & I40E_DEBUG_USER)
6578                 dev_info(&pf->pdev->dev,
6579                          "pf=%d, num_vfs=%d, msix_pf=%d, msix_vf=%d, fd_g=%d, fd_b=%d, pf_max_q=%d num_vsi=%d\n",
6580                          pf->hw.pf_id, pf->hw.func_caps.num_vfs,
6581                          pf->hw.func_caps.num_msix_vectors,
6582                          pf->hw.func_caps.num_msix_vectors_vf,
6583                          pf->hw.func_caps.fd_filters_guaranteed,
6584                          pf->hw.func_caps.fd_filters_best_effort,
6585                          pf->hw.func_caps.num_tx_qp,
6586                          pf->hw.func_caps.num_vsis);
6587
6588 #define DEF_NUM_VSI (1 + (pf->hw.func_caps.fcoe ? 1 : 0) \
6589                        + pf->hw.func_caps.num_vfs)
6590         if (pf->hw.revision_id == 0 && (DEF_NUM_VSI > pf->hw.func_caps.num_vsis)) {
6591                 dev_info(&pf->pdev->dev,
6592                          "got num_vsis %d, setting num_vsis to %d\n",
6593                          pf->hw.func_caps.num_vsis, DEF_NUM_VSI);
6594                 pf->hw.func_caps.num_vsis = DEF_NUM_VSI;
6595         }
6596
6597         return 0;
6598 }
6599
6600 static int i40e_vsi_clear(struct i40e_vsi *vsi);
6601
6602 /**
6603  * i40e_fdir_sb_setup - initialize the Flow Director resources for Sideband
6604  * @pf: board private structure
6605  **/
6606 static void i40e_fdir_sb_setup(struct i40e_pf *pf)
6607 {
6608         struct i40e_vsi *vsi;
6609         int i;
6610
6611         /* quick workaround for an NVM issue that leaves a critical register
6612          * uninitialized
6613          */
6614         if (!rd32(&pf->hw, I40E_GLQF_HKEY(0))) {
6615                 static const u32 hkey[] = {
6616                         0xe640d33f, 0xcdfe98ab, 0x73fa7161, 0x0d7a7d36,
6617                         0xeacb7d61, 0xaa4f05b6, 0x9c5c89ed, 0xfc425ddb,
6618                         0xa4654832, 0xfc7461d4, 0x8f827619, 0xf5c63c21,
6619                         0x95b3a76d};
6620
6621                 for (i = 0; i <= I40E_GLQF_HKEY_MAX_INDEX; i++)
6622                         wr32(&pf->hw, I40E_GLQF_HKEY(i), hkey[i]);
6623         }
6624
6625         if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
6626                 return;
6627
6628         /* find existing VSI and see if it needs configuring */
6629         vsi = NULL;
6630         for (i = 0; i < pf->num_alloc_vsi; i++) {
6631                 if (pf->vsi[i] && pf->vsi[i]->type == I40E_VSI_FDIR) {
6632                         vsi = pf->vsi[i];
6633                         break;
6634                 }
6635         }
6636
6637         /* create a new VSI if none exists */
6638         if (!vsi) {
6639                 vsi = i40e_vsi_setup(pf, I40E_VSI_FDIR,
6640                                      pf->vsi[pf->lan_vsi]->seid, 0);
6641                 if (!vsi) {
6642                         dev_info(&pf->pdev->dev, "Couldn't create FDir VSI\n");
6643                         pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
6644                         return;
6645                 }
6646         }
6647
6648         i40e_vsi_setup_irqhandler(vsi, i40e_fdir_clean_ring);
6649 }
6650
6651 /**
6652  * i40e_fdir_teardown - release the Flow Director resources
6653  * @pf: board private structure
6654  **/
6655 static void i40e_fdir_teardown(struct i40e_pf *pf)
6656 {
6657         int i;
6658
6659         i40e_fdir_filter_exit(pf);
6660         for (i = 0; i < pf->num_alloc_vsi; i++) {
6661                 if (pf->vsi[i] && pf->vsi[i]->type == I40E_VSI_FDIR) {
6662                         i40e_vsi_release(pf->vsi[i]);
6663                         break;
6664                 }
6665         }
6666 }
6667
6668 /**
6669  * i40e_prep_for_reset - prep for the core to reset
6670  * @pf: board private structure
6671  *
6672  * Close up the VFs and other things in prep for PF Reset.
6673   **/
6674 static void i40e_prep_for_reset(struct i40e_pf *pf)
6675 {
6676         struct i40e_hw *hw = &pf->hw;
6677         i40e_status ret = 0;
6678         u32 v;
6679
6680         clear_bit(__I40E_RESET_INTR_RECEIVED, &pf->state);
6681         if (test_and_set_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state))
6682                 return;
6683
6684         dev_dbg(&pf->pdev->dev, "Tearing down internal switch for reset\n");
6685
6686         /* quiesce the VSIs and their queues that are not already DOWN */
6687         i40e_pf_quiesce_all_vsi(pf);
6688
6689         for (v = 0; v < pf->num_alloc_vsi; v++) {
6690                 if (pf->vsi[v])
6691                         pf->vsi[v]->seid = 0;
6692         }
6693
6694         i40e_shutdown_adminq(&pf->hw);
6695
6696         /* call shutdown HMC */
6697         if (hw->hmc.hmc_obj) {
6698                 ret = i40e_shutdown_lan_hmc(hw);
6699                 if (ret)
6700                         dev_warn(&pf->pdev->dev,
6701                                  "shutdown_lan_hmc failed: %d\n", ret);
6702         }
6703 }
6704
6705 /**
6706  * i40e_send_version - update firmware with driver version
6707  * @pf: PF struct
6708  */
6709 static void i40e_send_version(struct i40e_pf *pf)
6710 {
6711         struct i40e_driver_version dv;
6712
6713         dv.major_version = DRV_VERSION_MAJOR;
6714         dv.minor_version = DRV_VERSION_MINOR;
6715         dv.build_version = DRV_VERSION_BUILD;
6716         dv.subbuild_version = 0;
6717         strlcpy(dv.driver_string, DRV_VERSION, sizeof(dv.driver_string));
6718         i40e_aq_send_driver_version(&pf->hw, &dv, NULL);
6719 }
6720
6721 /**
6722  * i40e_reset_and_rebuild - reset and rebuild using a saved config
6723  * @pf: board private structure
6724  * @reinit: if the Main VSI needs to re-initialized.
6725  **/
6726 static void i40e_reset_and_rebuild(struct i40e_pf *pf, bool reinit)
6727 {
6728         struct i40e_hw *hw = &pf->hw;
6729         u8 set_fc_aq_fail = 0;
6730         i40e_status ret;
6731         u32 val;
6732         u32 v;
6733
6734         /* Now we wait for GRST to settle out.
6735          * We don't have to delete the VEBs or VSIs from the hw switch
6736          * because the reset will make them disappear.
6737          */
6738         ret = i40e_pf_reset(hw);
6739         if (ret) {
6740                 dev_info(&pf->pdev->dev, "PF reset failed, %d\n", ret);
6741                 set_bit(__I40E_RESET_FAILED, &pf->state);
6742                 goto clear_recovery;
6743         }
6744         pf->pfr_count++;
6745
6746         if (test_bit(__I40E_DOWN, &pf->state))
6747                 goto clear_recovery;
6748         dev_dbg(&pf->pdev->dev, "Rebuilding internal switch\n");
6749
6750         /* rebuild the basics for the AdminQ, HMC, and initial HW switch */
6751         ret = i40e_init_adminq(&pf->hw);
6752         if (ret) {
6753                 dev_info(&pf->pdev->dev, "Rebuild AdminQ failed, err %s aq_err %s\n",
6754                          i40e_stat_str(&pf->hw, ret),
6755                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6756                 goto clear_recovery;
6757         }
6758
6759         /* re-verify the eeprom if we just had an EMP reset */
6760         if (test_and_clear_bit(__I40E_EMP_RESET_INTR_RECEIVED, &pf->state))
6761                 i40e_verify_eeprom(pf);
6762
6763         i40e_clear_pxe_mode(hw);
6764         ret = i40e_get_capabilities(pf);
6765         if (ret)
6766                 goto end_core_reset;
6767
6768         ret = i40e_init_lan_hmc(hw, hw->func_caps.num_tx_qp,
6769                                 hw->func_caps.num_rx_qp,
6770                                 pf->fcoe_hmc_cntx_num, pf->fcoe_hmc_filt_num);
6771         if (ret) {
6772                 dev_info(&pf->pdev->dev, "init_lan_hmc failed: %d\n", ret);
6773                 goto end_core_reset;
6774         }
6775         ret = i40e_configure_lan_hmc(hw, I40E_HMC_MODEL_DIRECT_ONLY);
6776         if (ret) {
6777                 dev_info(&pf->pdev->dev, "configure_lan_hmc failed: %d\n", ret);
6778                 goto end_core_reset;
6779         }
6780
6781 #ifdef CONFIG_I40E_DCB
6782         ret = i40e_init_pf_dcb(pf);
6783         if (ret) {
6784                 dev_info(&pf->pdev->dev, "DCB init failed %d, disabled\n", ret);
6785                 pf->flags &= ~I40E_FLAG_DCB_CAPABLE;
6786                 /* Continue without DCB enabled */
6787         }
6788 #endif /* CONFIG_I40E_DCB */
6789 #ifdef I40E_FCOE
6790         i40e_init_pf_fcoe(pf);
6791
6792 #endif
6793         /* do basic switch setup */
6794         ret = i40e_setup_pf_switch(pf, reinit);
6795         if (ret)
6796                 goto end_core_reset;
6797
6798         /* driver is only interested in link up/down and module qualification
6799          * reports from firmware
6800          */
6801         ret = i40e_aq_set_phy_int_mask(&pf->hw,
6802                                        I40E_AQ_EVENT_LINK_UPDOWN |
6803                                        I40E_AQ_EVENT_MODULE_QUAL_FAIL, NULL);
6804         if (ret)
6805                 dev_info(&pf->pdev->dev, "set phy mask fail, err %s aq_err %s\n",
6806                          i40e_stat_str(&pf->hw, ret),
6807                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6808
6809         /* make sure our flow control settings are restored */
6810         ret = i40e_set_fc(&pf->hw, &set_fc_aq_fail, true);
6811         if (ret)
6812                 dev_dbg(&pf->pdev->dev, "setting flow control: ret = %s last_status = %s\n",
6813                         i40e_stat_str(&pf->hw, ret),
6814                         i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6815
6816         /* Rebuild the VSIs and VEBs that existed before reset.
6817          * They are still in our local switch element arrays, so only
6818          * need to rebuild the switch model in the HW.
6819          *
6820          * If there were VEBs but the reconstitution failed, we'll try
6821          * try to recover minimal use by getting the basic PF VSI working.
6822          */
6823         if (pf->vsi[pf->lan_vsi]->uplink_seid != pf->mac_seid) {
6824                 dev_dbg(&pf->pdev->dev, "attempting to rebuild switch\n");
6825                 /* find the one VEB connected to the MAC, and find orphans */
6826                 for (v = 0; v < I40E_MAX_VEB; v++) {
6827                         if (!pf->veb[v])
6828                                 continue;
6829
6830                         if (pf->veb[v]->uplink_seid == pf->mac_seid ||
6831                             pf->veb[v]->uplink_seid == 0) {
6832                                 ret = i40e_reconstitute_veb(pf->veb[v]);
6833
6834                                 if (!ret)
6835                                         continue;
6836
6837                                 /* If Main VEB failed, we're in deep doodoo,
6838                                  * so give up rebuilding the switch and set up
6839                                  * for minimal rebuild of PF VSI.
6840                                  * If orphan failed, we'll report the error
6841                                  * but try to keep going.
6842                                  */
6843                                 if (pf->veb[v]->uplink_seid == pf->mac_seid) {
6844                                         dev_info(&pf->pdev->dev,
6845                                                  "rebuild of switch failed: %d, will try to set up simple PF connection\n",
6846                                                  ret);
6847                                         pf->vsi[pf->lan_vsi]->uplink_seid
6848                                                                 = pf->mac_seid;
6849                                         break;
6850                                 } else if (pf->veb[v]->uplink_seid == 0) {
6851                                         dev_info(&pf->pdev->dev,
6852                                                  "rebuild of orphan VEB failed: %d\n",
6853                                                  ret);
6854                                 }
6855                         }
6856                 }
6857         }
6858
6859         if (pf->vsi[pf->lan_vsi]->uplink_seid == pf->mac_seid) {
6860                 dev_dbg(&pf->pdev->dev, "attempting to rebuild PF VSI\n");
6861                 /* no VEB, so rebuild only the Main VSI */
6862                 ret = i40e_add_vsi(pf->vsi[pf->lan_vsi]);
6863                 if (ret) {
6864                         dev_info(&pf->pdev->dev,
6865                                  "rebuild of Main VSI failed: %d\n", ret);
6866                         goto end_core_reset;
6867                 }
6868         }
6869
6870         /* Reconfigure hardware for allowing smaller MSS in the case
6871          * of TSO, so that we avoid the MDD being fired and causing
6872          * a reset in the case of small MSS+TSO.
6873          */
6874 #define I40E_REG_MSS          0x000E64DC
6875 #define I40E_REG_MSS_MIN_MASK 0x3FF0000
6876 #define I40E_64BYTE_MSS       0x400000
6877         val = rd32(hw, I40E_REG_MSS);
6878         if ((val & I40E_REG_MSS_MIN_MASK) > I40E_64BYTE_MSS) {
6879                 val &= ~I40E_REG_MSS_MIN_MASK;
6880                 val |= I40E_64BYTE_MSS;
6881                 wr32(hw, I40E_REG_MSS, val);
6882         }
6883
6884         if (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver < 33)) ||
6885             (pf->hw.aq.fw_maj_ver < 4)) {
6886                 msleep(75);
6887                 ret = i40e_aq_set_link_restart_an(&pf->hw, true, NULL);
6888                 if (ret)
6889                         dev_info(&pf->pdev->dev, "link restart failed, err %s aq_err %s\n",
6890                                  i40e_stat_str(&pf->hw, ret),
6891                                  i40e_aq_str(&pf->hw,
6892                                              pf->hw.aq.asq_last_status));
6893         }
6894         /* reinit the misc interrupt */
6895         if (pf->flags & I40E_FLAG_MSIX_ENABLED)
6896                 ret = i40e_setup_misc_vector(pf);
6897
6898         /* Add a filter to drop all Flow control frames from any VSI from being
6899          * transmitted. By doing so we stop a malicious VF from sending out
6900          * PAUSE or PFC frames and potentially controlling traffic for other
6901          * PF/VF VSIs.
6902          * The FW can still send Flow control frames if enabled.
6903          */
6904         i40e_add_filter_to_drop_tx_flow_control_frames(&pf->hw,
6905                                                        pf->main_vsi_seid);
6906
6907         /* restart the VSIs that were rebuilt and running before the reset */
6908         i40e_pf_unquiesce_all_vsi(pf);
6909
6910         if (pf->num_alloc_vfs) {
6911                 for (v = 0; v < pf->num_alloc_vfs; v++)
6912                         i40e_reset_vf(&pf->vf[v], true);
6913         }
6914
6915         /* tell the firmware that we're starting */
6916         i40e_send_version(pf);
6917
6918 end_core_reset:
6919         clear_bit(__I40E_RESET_FAILED, &pf->state);
6920 clear_recovery:
6921         clear_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state);
6922 }
6923
6924 /**
6925  * i40e_handle_reset_warning - prep for the PF to reset, reset and rebuild
6926  * @pf: board private structure
6927  *
6928  * Close up the VFs and other things in prep for a Core Reset,
6929  * then get ready to rebuild the world.
6930  **/
6931 static void i40e_handle_reset_warning(struct i40e_pf *pf)
6932 {
6933         i40e_prep_for_reset(pf);
6934         i40e_reset_and_rebuild(pf, false);
6935 }
6936
6937 /**
6938  * i40e_handle_mdd_event
6939  * @pf: pointer to the PF structure
6940  *
6941  * Called from the MDD irq handler to identify possibly malicious vfs
6942  **/
6943 static void i40e_handle_mdd_event(struct i40e_pf *pf)
6944 {
6945         struct i40e_hw *hw = &pf->hw;
6946         bool mdd_detected = false;
6947         bool pf_mdd_detected = false;
6948         struct i40e_vf *vf;
6949         u32 reg;
6950         int i;
6951
6952         if (!test_bit(__I40E_MDD_EVENT_PENDING, &pf->state))
6953                 return;
6954
6955         /* find what triggered the MDD event */
6956         reg = rd32(hw, I40E_GL_MDET_TX);
6957         if (reg & I40E_GL_MDET_TX_VALID_MASK) {
6958                 u8 pf_num = (reg & I40E_GL_MDET_TX_PF_NUM_MASK) >>
6959                                 I40E_GL_MDET_TX_PF_NUM_SHIFT;
6960                 u16 vf_num = (reg & I40E_GL_MDET_TX_VF_NUM_MASK) >>
6961                                 I40E_GL_MDET_TX_VF_NUM_SHIFT;
6962                 u8 event = (reg & I40E_GL_MDET_TX_EVENT_MASK) >>
6963                                 I40E_GL_MDET_TX_EVENT_SHIFT;
6964                 u16 queue = ((reg & I40E_GL_MDET_TX_QUEUE_MASK) >>
6965                                 I40E_GL_MDET_TX_QUEUE_SHIFT) -
6966                                 pf->hw.func_caps.base_queue;
6967                 if (netif_msg_tx_err(pf))
6968                         dev_info(&pf->pdev->dev, "Malicious Driver Detection event 0x%02x on TX queue %d PF number 0x%02x VF number 0x%02x\n",
6969                                  event, queue, pf_num, vf_num);
6970                 wr32(hw, I40E_GL_MDET_TX, 0xffffffff);
6971                 mdd_detected = true;
6972         }
6973         reg = rd32(hw, I40E_GL_MDET_RX);
6974         if (reg & I40E_GL_MDET_RX_VALID_MASK) {
6975                 u8 func = (reg & I40E_GL_MDET_RX_FUNCTION_MASK) >>
6976                                 I40E_GL_MDET_RX_FUNCTION_SHIFT;
6977                 u8 event = (reg & I40E_GL_MDET_RX_EVENT_MASK) >>
6978                                 I40E_GL_MDET_RX_EVENT_SHIFT;
6979                 u16 queue = ((reg & I40E_GL_MDET_RX_QUEUE_MASK) >>
6980                                 I40E_GL_MDET_RX_QUEUE_SHIFT) -
6981                                 pf->hw.func_caps.base_queue;
6982                 if (netif_msg_rx_err(pf))
6983                         dev_info(&pf->pdev->dev, "Malicious Driver Detection event 0x%02x on RX queue %d of function 0x%02x\n",
6984                                  event, queue, func);
6985                 wr32(hw, I40E_GL_MDET_RX, 0xffffffff);
6986                 mdd_detected = true;
6987         }
6988
6989         if (mdd_detected) {
6990                 reg = rd32(hw, I40E_PF_MDET_TX);
6991                 if (reg & I40E_PF_MDET_TX_VALID_MASK) {
6992                         wr32(hw, I40E_PF_MDET_TX, 0xFFFF);
6993                         dev_info(&pf->pdev->dev, "TX driver issue detected, PF reset issued\n");
6994                         pf_mdd_detected = true;
6995                 }
6996                 reg = rd32(hw, I40E_PF_MDET_RX);
6997                 if (reg & I40E_PF_MDET_RX_VALID_MASK) {
6998                         wr32(hw, I40E_PF_MDET_RX, 0xFFFF);
6999                         dev_info(&pf->pdev->dev, "RX driver issue detected, PF reset issued\n");
7000                         pf_mdd_detected = true;
7001                 }
7002                 /* Queue belongs to the PF, initiate a reset */
7003                 if (pf_mdd_detected) {
7004                         set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
7005                         i40e_service_event_schedule(pf);
7006                 }
7007         }
7008
7009         /* see if one of the VFs needs its hand slapped */
7010         for (i = 0; i < pf->num_alloc_vfs && mdd_detected; i++) {
7011                 vf = &(pf->vf[i]);
7012                 reg = rd32(hw, I40E_VP_MDET_TX(i));
7013                 if (reg & I40E_VP_MDET_TX_VALID_MASK) {
7014                         wr32(hw, I40E_VP_MDET_TX(i), 0xFFFF);
7015                         vf->num_mdd_events++;
7016                         dev_info(&pf->pdev->dev, "TX driver issue detected on VF %d\n",
7017                                  i);
7018                 }
7019
7020                 reg = rd32(hw, I40E_VP_MDET_RX(i));
7021                 if (reg & I40E_VP_MDET_RX_VALID_MASK) {
7022                         wr32(hw, I40E_VP_MDET_RX(i), 0xFFFF);
7023                         vf->num_mdd_events++;
7024                         dev_info(&pf->pdev->dev, "RX driver issue detected on VF %d\n",
7025                                  i);
7026                 }
7027
7028                 if (vf->num_mdd_events > I40E_DEFAULT_NUM_MDD_EVENTS_ALLOWED) {
7029                         dev_info(&pf->pdev->dev,
7030                                  "Too many MDD events on VF %d, disabled\n", i);
7031                         dev_info(&pf->pdev->dev,
7032                                  "Use PF Control I/F to re-enable the VF\n");
7033                         set_bit(I40E_VF_STAT_DISABLED, &vf->vf_states);
7034                 }
7035         }
7036
7037         /* re-enable mdd interrupt cause */
7038         clear_bit(__I40E_MDD_EVENT_PENDING, &pf->state);
7039         reg = rd32(hw, I40E_PFINT_ICR0_ENA);
7040         reg |=  I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK;
7041         wr32(hw, I40E_PFINT_ICR0_ENA, reg);
7042         i40e_flush(hw);
7043 }
7044
7045 /**
7046  * i40e_sync_udp_filters_subtask - Sync the VSI filter list with HW
7047  * @pf: board private structure
7048  **/
7049 static void i40e_sync_udp_filters_subtask(struct i40e_pf *pf)
7050 {
7051 #if IS_ENABLED(CONFIG_VXLAN) || IS_ENABLED(CONFIG_GENEVE)
7052         struct i40e_hw *hw = &pf->hw;
7053         i40e_status ret;
7054         __be16 port;
7055         int i;
7056
7057         if (!(pf->flags & I40E_FLAG_UDP_FILTER_SYNC))
7058                 return;
7059
7060         pf->flags &= ~I40E_FLAG_UDP_FILTER_SYNC;
7061
7062         for (i = 0; i < I40E_MAX_PF_UDP_OFFLOAD_PORTS; i++) {
7063                 if (pf->pending_udp_bitmap & BIT_ULL(i)) {
7064                         pf->pending_udp_bitmap &= ~BIT_ULL(i);
7065                         port = pf->udp_ports[i].index;
7066                         if (port)
7067                                 ret = i40e_aq_add_udp_tunnel(hw, ntohs(port),
7068                                                      pf->udp_ports[i].type,
7069                                                      NULL, NULL);
7070                         else
7071                                 ret = i40e_aq_del_udp_tunnel(hw, i, NULL);
7072
7073                         if (ret) {
7074                                 dev_info(&pf->pdev->dev,
7075                                          "%s vxlan port %d, index %d failed, err %s aq_err %s\n",
7076                                          port ? "add" : "delete",
7077                                          ntohs(port), i,
7078                                          i40e_stat_str(&pf->hw, ret),
7079                                          i40e_aq_str(&pf->hw,
7080                                                     pf->hw.aq.asq_last_status));
7081                                 pf->udp_ports[i].index = 0;
7082                         }
7083                 }
7084         }
7085 #endif
7086 }
7087
7088 /**
7089  * i40e_service_task - Run the driver's async subtasks
7090  * @work: pointer to work_struct containing our data
7091  **/
7092 static void i40e_service_task(struct work_struct *work)
7093 {
7094         struct i40e_pf *pf = container_of(work,
7095                                           struct i40e_pf,
7096                                           service_task);
7097         unsigned long start_time = jiffies;
7098
7099         /* don't bother with service tasks if a reset is in progress */
7100         if (test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state)) {
7101                 i40e_service_event_complete(pf);
7102                 return;
7103         }
7104
7105         i40e_detect_recover_hung(pf);
7106         i40e_reset_subtask(pf);
7107         i40e_handle_mdd_event(pf);
7108         i40e_vc_process_vflr_event(pf);
7109         i40e_watchdog_subtask(pf);
7110         i40e_fdir_reinit_subtask(pf);
7111         i40e_sync_filters_subtask(pf);
7112 #if IS_ENABLED(CONFIG_VXLAN) || IS_ENABLED(CONFIG_GENEVE)
7113         i40e_sync_udp_filters_subtask(pf);
7114 #endif
7115         i40e_clean_adminq_subtask(pf);
7116
7117         i40e_service_event_complete(pf);
7118
7119         /* If the tasks have taken longer than one timer cycle or there
7120          * is more work to be done, reschedule the service task now
7121          * rather than wait for the timer to tick again.
7122          */
7123         if (time_after(jiffies, (start_time + pf->service_timer_period)) ||
7124             test_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state)            ||
7125             test_bit(__I40E_MDD_EVENT_PENDING, &pf->state)               ||
7126             test_bit(__I40E_VFLR_EVENT_PENDING, &pf->state))
7127                 i40e_service_event_schedule(pf);
7128 }
7129
7130 /**
7131  * i40e_service_timer - timer callback
7132  * @data: pointer to PF struct
7133  **/
7134 static void i40e_service_timer(unsigned long data)
7135 {
7136         struct i40e_pf *pf = (struct i40e_pf *)data;
7137
7138         mod_timer(&pf->service_timer,
7139                   round_jiffies(jiffies + pf->service_timer_period));
7140         i40e_service_event_schedule(pf);
7141 }
7142
7143 /**
7144  * i40e_set_num_rings_in_vsi - Determine number of rings in the VSI
7145  * @vsi: the VSI being configured
7146  **/
7147 static int i40e_set_num_rings_in_vsi(struct i40e_vsi *vsi)
7148 {
7149         struct i40e_pf *pf = vsi->back;
7150
7151         switch (vsi->type) {
7152         case I40E_VSI_MAIN:
7153                 vsi->alloc_queue_pairs = pf->num_lan_qps;
7154                 vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
7155                                       I40E_REQ_DESCRIPTOR_MULTIPLE);
7156                 if (pf->flags & I40E_FLAG_MSIX_ENABLED)
7157                         vsi->num_q_vectors = pf->num_lan_msix;
7158                 else
7159                         vsi->num_q_vectors = 1;
7160
7161                 break;
7162
7163         case I40E_VSI_FDIR:
7164                 vsi->alloc_queue_pairs = 1;
7165                 vsi->num_desc = ALIGN(I40E_FDIR_RING_COUNT,
7166                                       I40E_REQ_DESCRIPTOR_MULTIPLE);
7167                 vsi->num_q_vectors = 1;
7168                 break;
7169
7170         case I40E_VSI_VMDQ2:
7171                 vsi->alloc_queue_pairs = pf->num_vmdq_qps;
7172                 vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
7173                                       I40E_REQ_DESCRIPTOR_MULTIPLE);
7174                 vsi->num_q_vectors = pf->num_vmdq_msix;
7175                 break;
7176
7177         case I40E_VSI_SRIOV:
7178                 vsi->alloc_queue_pairs = pf->num_vf_qps;
7179                 vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
7180                                       I40E_REQ_DESCRIPTOR_MULTIPLE);
7181                 break;
7182
7183 #ifdef I40E_FCOE
7184         case I40E_VSI_FCOE:
7185                 vsi->alloc_queue_pairs = pf->num_fcoe_qps;
7186                 vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
7187                                       I40E_REQ_DESCRIPTOR_MULTIPLE);
7188                 vsi->num_q_vectors = pf->num_fcoe_msix;
7189                 break;
7190
7191 #endif /* I40E_FCOE */
7192         default:
7193                 WARN_ON(1);
7194                 return -ENODATA;
7195         }
7196
7197         return 0;
7198 }
7199
7200 /**
7201  * i40e_vsi_alloc_arrays - Allocate queue and vector pointer arrays for the vsi
7202  * @type: VSI pointer
7203  * @alloc_qvectors: a bool to specify if q_vectors need to be allocated.
7204  *
7205  * On error: returns error code (negative)
7206  * On success: returns 0
7207  **/
7208 static int i40e_vsi_alloc_arrays(struct i40e_vsi *vsi, bool alloc_qvectors)
7209 {
7210         int size;
7211         int ret = 0;
7212
7213         /* allocate memory for both Tx and Rx ring pointers */
7214         size = sizeof(struct i40e_ring *) * vsi->alloc_queue_pairs * 2;
7215         vsi->tx_rings = kzalloc(size, GFP_KERNEL);
7216         if (!vsi->tx_rings)
7217                 return -ENOMEM;
7218         vsi->rx_rings = &vsi->tx_rings[vsi->alloc_queue_pairs];
7219
7220         if (alloc_qvectors) {
7221                 /* allocate memory for q_vector pointers */
7222                 size = sizeof(struct i40e_q_vector *) * vsi->num_q_vectors;
7223                 vsi->q_vectors = kzalloc(size, GFP_KERNEL);
7224                 if (!vsi->q_vectors) {
7225                         ret = -ENOMEM;
7226                         goto err_vectors;
7227                 }
7228         }
7229         return ret;
7230
7231 err_vectors:
7232         kfree(vsi->tx_rings);
7233         return ret;
7234 }
7235
7236 /**
7237  * i40e_vsi_mem_alloc - Allocates the next available struct vsi in the PF
7238  * @pf: board private structure
7239  * @type: type of VSI
7240  *
7241  * On error: returns error code (negative)
7242  * On success: returns vsi index in PF (positive)
7243  **/
7244 static int i40e_vsi_mem_alloc(struct i40e_pf *pf, enum i40e_vsi_type type)
7245 {
7246         int ret = -ENODEV;
7247         struct i40e_vsi *vsi;
7248         int vsi_idx;
7249         int i;
7250
7251         /* Need to protect the allocation of the VSIs at the PF level */
7252         mutex_lock(&pf->switch_mutex);
7253
7254         /* VSI list may be fragmented if VSI creation/destruction has
7255          * been happening.  We can afford to do a quick scan to look
7256          * for any free VSIs in the list.
7257          *
7258          * find next empty vsi slot, looping back around if necessary
7259          */
7260         i = pf->next_vsi;
7261         while (i < pf->num_alloc_vsi && pf->vsi[i])
7262                 i++;
7263         if (i >= pf->num_alloc_vsi) {
7264                 i = 0;
7265                 while (i < pf->next_vsi && pf->vsi[i])
7266                         i++;
7267         }
7268
7269         if (i < pf->num_alloc_vsi && !pf->vsi[i]) {
7270                 vsi_idx = i;             /* Found one! */
7271         } else {
7272                 ret = -ENODEV;
7273                 goto unlock_pf;  /* out of VSI slots! */
7274         }
7275         pf->next_vsi = ++i;
7276
7277         vsi = kzalloc(sizeof(*vsi), GFP_KERNEL);
7278         if (!vsi) {
7279                 ret = -ENOMEM;
7280                 goto unlock_pf;
7281         }
7282         vsi->type = type;
7283         vsi->back = pf;
7284         set_bit(__I40E_DOWN, &vsi->state);
7285         vsi->flags = 0;
7286         vsi->idx = vsi_idx;
7287         vsi->rx_itr_setting = pf->rx_itr_default;
7288         vsi->tx_itr_setting = pf->tx_itr_default;
7289         vsi->int_rate_limit = 0;
7290         vsi->rss_table_size = (vsi->type == I40E_VSI_MAIN) ?
7291                                 pf->rss_table_size : 64;
7292         vsi->netdev_registered = false;
7293         vsi->work_limit = I40E_DEFAULT_IRQ_WORK;
7294         INIT_LIST_HEAD(&vsi->mac_filter_list);
7295         vsi->irqs_ready = false;
7296
7297         ret = i40e_set_num_rings_in_vsi(vsi);
7298         if (ret)
7299                 goto err_rings;
7300
7301         ret = i40e_vsi_alloc_arrays(vsi, true);
7302         if (ret)
7303                 goto err_rings;
7304
7305         /* Setup default MSIX irq handler for VSI */
7306         i40e_vsi_setup_irqhandler(vsi, i40e_msix_clean_rings);
7307
7308         /* Initialize VSI lock */
7309         spin_lock_init(&vsi->mac_filter_list_lock);
7310         pf->vsi[vsi_idx] = vsi;
7311         ret = vsi_idx;
7312         goto unlock_pf;
7313
7314 err_rings:
7315         pf->next_vsi = i - 1;
7316         kfree(vsi);
7317 unlock_pf:
7318         mutex_unlock(&pf->switch_mutex);
7319         return ret;
7320 }
7321
7322 /**
7323  * i40e_vsi_free_arrays - Free queue and vector pointer arrays for the VSI
7324  * @type: VSI pointer
7325  * @free_qvectors: a bool to specify if q_vectors need to be freed.
7326  *
7327  * On error: returns error code (negative)
7328  * On success: returns 0
7329  **/
7330 static void i40e_vsi_free_arrays(struct i40e_vsi *vsi, bool free_qvectors)
7331 {
7332         /* free the ring and vector containers */
7333         if (free_qvectors) {
7334                 kfree(vsi->q_vectors);
7335                 vsi->q_vectors = NULL;
7336         }
7337         kfree(vsi->tx_rings);
7338         vsi->tx_rings = NULL;
7339         vsi->rx_rings = NULL;
7340 }
7341
7342 /**
7343  * i40e_clear_rss_config_user - clear the user configured RSS hash keys
7344  * and lookup table
7345  * @vsi: Pointer to VSI structure
7346  */
7347 static void i40e_clear_rss_config_user(struct i40e_vsi *vsi)
7348 {
7349         if (!vsi)
7350                 return;
7351
7352         kfree(vsi->rss_hkey_user);
7353         vsi->rss_hkey_user = NULL;
7354
7355         kfree(vsi->rss_lut_user);
7356         vsi->rss_lut_user = NULL;
7357 }
7358
7359 /**
7360  * i40e_vsi_clear - Deallocate the VSI provided
7361  * @vsi: the VSI being un-configured
7362  **/
7363 static int i40e_vsi_clear(struct i40e_vsi *vsi)
7364 {
7365         struct i40e_pf *pf;
7366
7367         if (!vsi)
7368                 return 0;
7369
7370         if (!vsi->back)
7371                 goto free_vsi;
7372         pf = vsi->back;
7373
7374         mutex_lock(&pf->switch_mutex);
7375         if (!pf->vsi[vsi->idx]) {
7376                 dev_err(&pf->pdev->dev, "pf->vsi[%d] is NULL, just free vsi[%d](%p,type %d)\n",
7377                         vsi->idx, vsi->idx, vsi, vsi->type);
7378                 goto unlock_vsi;
7379         }
7380
7381         if (pf->vsi[vsi->idx] != vsi) {
7382                 dev_err(&pf->pdev->dev,
7383                         "pf->vsi[%d](%p, type %d) != vsi[%d](%p,type %d): no free!\n",
7384                         pf->vsi[vsi->idx]->idx,
7385                         pf->vsi[vsi->idx],
7386                         pf->vsi[vsi->idx]->type,
7387                         vsi->idx, vsi, vsi->type);
7388                 goto unlock_vsi;
7389         }
7390
7391         /* updates the PF for this cleared vsi */
7392         i40e_put_lump(pf->qp_pile, vsi->base_queue, vsi->idx);
7393         i40e_put_lump(pf->irq_pile, vsi->base_vector, vsi->idx);
7394
7395         i40e_vsi_free_arrays(vsi, true);
7396         i40e_clear_rss_config_user(vsi);
7397
7398         pf->vsi[vsi->idx] = NULL;
7399         if (vsi->idx < pf->next_vsi)
7400                 pf->next_vsi = vsi->idx;
7401
7402 unlock_vsi:
7403         mutex_unlock(&pf->switch_mutex);
7404 free_vsi:
7405         kfree(vsi);
7406
7407         return 0;
7408 }
7409
7410 /**
7411  * i40e_vsi_clear_rings - Deallocates the Rx and Tx rings for the provided VSI
7412  * @vsi: the VSI being cleaned
7413  **/
7414 static void i40e_vsi_clear_rings(struct i40e_vsi *vsi)
7415 {
7416         int i;
7417
7418         if (vsi->tx_rings && vsi->tx_rings[0]) {
7419                 for (i = 0; i < vsi->alloc_queue_pairs; i++) {
7420                         kfree_rcu(vsi->tx_rings[i], rcu);
7421                         vsi->tx_rings[i] = NULL;
7422                         vsi->rx_rings[i] = NULL;
7423                 }
7424         }
7425 }
7426
7427 /**
7428  * i40e_alloc_rings - Allocates the Rx and Tx rings for the provided VSI
7429  * @vsi: the VSI being configured
7430  **/
7431 static int i40e_alloc_rings(struct i40e_vsi *vsi)
7432 {
7433         struct i40e_ring *tx_ring, *rx_ring;
7434         struct i40e_pf *pf = vsi->back;
7435         int i;
7436
7437         /* Set basic values in the rings to be used later during open() */
7438         for (i = 0; i < vsi->alloc_queue_pairs; i++) {
7439                 /* allocate space for both Tx and Rx in one shot */
7440                 tx_ring = kzalloc(sizeof(struct i40e_ring) * 2, GFP_KERNEL);
7441                 if (!tx_ring)
7442                         goto err_out;
7443
7444                 tx_ring->queue_index = i;
7445                 tx_ring->reg_idx = vsi->base_queue + i;
7446                 tx_ring->ring_active = false;
7447                 tx_ring->vsi = vsi;
7448                 tx_ring->netdev = vsi->netdev;
7449                 tx_ring->dev = &pf->pdev->dev;
7450                 tx_ring->count = vsi->num_desc;
7451                 tx_ring->size = 0;
7452                 tx_ring->dcb_tc = 0;
7453                 if (vsi->back->flags & I40E_FLAG_WB_ON_ITR_CAPABLE)
7454                         tx_ring->flags = I40E_TXR_FLAGS_WB_ON_ITR;
7455                 if (vsi->back->flags & I40E_FLAG_OUTER_UDP_CSUM_CAPABLE)
7456                         tx_ring->flags |= I40E_TXR_FLAGS_OUTER_UDP_CSUM;
7457                 vsi->tx_rings[i] = tx_ring;
7458
7459                 rx_ring = &tx_ring[1];
7460                 rx_ring->queue_index = i;
7461                 rx_ring->reg_idx = vsi->base_queue + i;
7462                 rx_ring->ring_active = false;
7463                 rx_ring->vsi = vsi;
7464                 rx_ring->netdev = vsi->netdev;
7465                 rx_ring->dev = &pf->pdev->dev;
7466                 rx_ring->count = vsi->num_desc;
7467                 rx_ring->size = 0;
7468                 rx_ring->dcb_tc = 0;
7469                 if (pf->flags & I40E_FLAG_16BYTE_RX_DESC_ENABLED)
7470                         set_ring_16byte_desc_enabled(rx_ring);
7471                 else
7472                         clear_ring_16byte_desc_enabled(rx_ring);
7473                 vsi->rx_rings[i] = rx_ring;
7474         }
7475
7476         return 0;
7477
7478 err_out:
7479         i40e_vsi_clear_rings(vsi);
7480         return -ENOMEM;
7481 }
7482
7483 /**
7484  * i40e_reserve_msix_vectors - Reserve MSI-X vectors in the kernel
7485  * @pf: board private structure
7486  * @vectors: the number of MSI-X vectors to request
7487  *
7488  * Returns the number of vectors reserved, or error
7489  **/
7490 static int i40e_reserve_msix_vectors(struct i40e_pf *pf, int vectors)
7491 {
7492         vectors = pci_enable_msix_range(pf->pdev, pf->msix_entries,
7493                                         I40E_MIN_MSIX, vectors);
7494         if (vectors < 0) {
7495                 dev_info(&pf->pdev->dev,
7496                          "MSI-X vector reservation failed: %d\n", vectors);
7497                 vectors = 0;
7498         }
7499
7500         return vectors;
7501 }
7502
7503 /**
7504  * i40e_init_msix - Setup the MSIX capability
7505  * @pf: board private structure
7506  *
7507  * Work with the OS to set up the MSIX vectors needed.
7508  *
7509  * Returns the number of vectors reserved or negative on failure
7510  **/
7511 static int i40e_init_msix(struct i40e_pf *pf)
7512 {
7513         struct i40e_hw *hw = &pf->hw;
7514         int vectors_left;
7515         int v_budget, i;
7516         int v_actual;
7517
7518         if (!(pf->flags & I40E_FLAG_MSIX_ENABLED))
7519                 return -ENODEV;
7520
7521         /* The number of vectors we'll request will be comprised of:
7522          *   - Add 1 for "other" cause for Admin Queue events, etc.
7523          *   - The number of LAN queue pairs
7524          *      - Queues being used for RSS.
7525          *              We don't need as many as max_rss_size vectors.
7526          *              use rss_size instead in the calculation since that
7527          *              is governed by number of cpus in the system.
7528          *      - assumes symmetric Tx/Rx pairing
7529          *   - The number of VMDq pairs
7530 #ifdef I40E_FCOE
7531          *   - The number of FCOE qps.
7532 #endif
7533          * Once we count this up, try the request.
7534          *
7535          * If we can't get what we want, we'll simplify to nearly nothing
7536          * and try again.  If that still fails, we punt.
7537          */
7538         vectors_left = hw->func_caps.num_msix_vectors;
7539         v_budget = 0;
7540
7541         /* reserve one vector for miscellaneous handler */
7542         if (vectors_left) {
7543                 v_budget++;
7544                 vectors_left--;
7545         }
7546
7547         /* reserve vectors for the main PF traffic queues */
7548         pf->num_lan_msix = min_t(int, num_online_cpus(), vectors_left);
7549         vectors_left -= pf->num_lan_msix;
7550         v_budget += pf->num_lan_msix;
7551
7552         /* reserve one vector for sideband flow director */
7553         if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
7554                 if (vectors_left) {
7555                         v_budget++;
7556                         vectors_left--;
7557                 } else {
7558                         pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
7559                 }
7560         }
7561
7562 #ifdef I40E_FCOE
7563         /* can we reserve enough for FCoE? */
7564         if (pf->flags & I40E_FLAG_FCOE_ENABLED) {
7565                 if (!vectors_left)
7566                         pf->num_fcoe_msix = 0;
7567                 else if (vectors_left >= pf->num_fcoe_qps)
7568                         pf->num_fcoe_msix = pf->num_fcoe_qps;
7569                 else
7570                         pf->num_fcoe_msix = 1;
7571                 v_budget += pf->num_fcoe_msix;
7572                 vectors_left -= pf->num_fcoe_msix;
7573         }
7574
7575 #endif
7576         /* any vectors left over go for VMDq support */
7577         if (pf->flags & I40E_FLAG_VMDQ_ENABLED) {
7578                 int vmdq_vecs_wanted = pf->num_vmdq_vsis * pf->num_vmdq_qps;
7579                 int vmdq_vecs = min_t(int, vectors_left, vmdq_vecs_wanted);
7580
7581                 /* if we're short on vectors for what's desired, we limit
7582                  * the queues per vmdq.  If this is still more than are
7583                  * available, the user will need to change the number of
7584                  * queues/vectors used by the PF later with the ethtool
7585                  * channels command
7586                  */
7587                 if (vmdq_vecs < vmdq_vecs_wanted)
7588                         pf->num_vmdq_qps = 1;
7589                 pf->num_vmdq_msix = pf->num_vmdq_qps;
7590
7591                 v_budget += vmdq_vecs;
7592                 vectors_left -= vmdq_vecs;
7593         }
7594
7595         pf->msix_entries = kcalloc(v_budget, sizeof(struct msix_entry),
7596                                    GFP_KERNEL);
7597         if (!pf->msix_entries)
7598                 return -ENOMEM;
7599
7600         for (i = 0; i < v_budget; i++)
7601                 pf->msix_entries[i].entry = i;
7602         v_actual = i40e_reserve_msix_vectors(pf, v_budget);
7603
7604         if (v_actual != v_budget) {
7605                 /* If we have limited resources, we will start with no vectors
7606                  * for the special features and then allocate vectors to some
7607                  * of these features based on the policy and at the end disable
7608                  * the features that did not get any vectors.
7609                  */
7610 #ifdef I40E_FCOE
7611                 pf->num_fcoe_qps = 0;
7612                 pf->num_fcoe_msix = 0;
7613 #endif
7614                 pf->num_vmdq_msix = 0;
7615         }
7616
7617         if (v_actual < I40E_MIN_MSIX) {
7618                 pf->flags &= ~I40E_FLAG_MSIX_ENABLED;
7619                 kfree(pf->msix_entries);
7620                 pf->msix_entries = NULL;
7621                 return -ENODEV;
7622
7623         } else if (v_actual == I40E_MIN_MSIX) {
7624                 /* Adjust for minimal MSIX use */
7625                 pf->num_vmdq_vsis = 0;
7626                 pf->num_vmdq_qps = 0;
7627                 pf->num_lan_qps = 1;
7628                 pf->num_lan_msix = 1;
7629
7630         } else if (v_actual != v_budget) {
7631                 int vec;
7632
7633                 /* reserve the misc vector */
7634                 vec = v_actual - 1;
7635
7636                 /* Scale vector usage down */
7637                 pf->num_vmdq_msix = 1;    /* force VMDqs to only one vector */
7638                 pf->num_vmdq_vsis = 1;
7639                 pf->num_vmdq_qps = 1;
7640                 pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
7641
7642                 /* partition out the remaining vectors */
7643                 switch (vec) {
7644                 case 2:
7645                         pf->num_lan_msix = 1;
7646                         break;
7647                 case 3:
7648 #ifdef I40E_FCOE
7649                         /* give one vector to FCoE */
7650                         if (pf->flags & I40E_FLAG_FCOE_ENABLED) {
7651                                 pf->num_lan_msix = 1;
7652                                 pf->num_fcoe_msix = 1;
7653                         }
7654 #else
7655                         pf->num_lan_msix = 2;
7656 #endif
7657                         break;
7658                 default:
7659 #ifdef I40E_FCOE
7660                         /* give one vector to FCoE */
7661                         if (pf->flags & I40E_FLAG_FCOE_ENABLED) {
7662                                 pf->num_fcoe_msix = 1;
7663                                 vec--;
7664                         }
7665 #endif
7666                         /* give the rest to the PF */
7667                         pf->num_lan_msix = min_t(int, vec, pf->num_lan_qps);
7668                         break;
7669                 }
7670         }
7671
7672         if ((pf->flags & I40E_FLAG_VMDQ_ENABLED) &&
7673             (pf->num_vmdq_msix == 0)) {
7674                 dev_info(&pf->pdev->dev, "VMDq disabled, not enough MSI-X vectors\n");
7675                 pf->flags &= ~I40E_FLAG_VMDQ_ENABLED;
7676         }
7677 #ifdef I40E_FCOE
7678
7679         if ((pf->flags & I40E_FLAG_FCOE_ENABLED) && (pf->num_fcoe_msix == 0)) {
7680                 dev_info(&pf->pdev->dev, "FCOE disabled, not enough MSI-X vectors\n");
7681                 pf->flags &= ~I40E_FLAG_FCOE_ENABLED;
7682         }
7683 #endif
7684         return v_actual;
7685 }
7686
7687 /**
7688  * i40e_vsi_alloc_q_vector - Allocate memory for a single interrupt vector
7689  * @vsi: the VSI being configured
7690  * @v_idx: index of the vector in the vsi struct
7691  *
7692  * We allocate one q_vector.  If allocation fails we return -ENOMEM.
7693  **/
7694 static int i40e_vsi_alloc_q_vector(struct i40e_vsi *vsi, int v_idx)
7695 {
7696         struct i40e_q_vector *q_vector;
7697
7698         /* allocate q_vector */
7699         q_vector = kzalloc(sizeof(struct i40e_q_vector), GFP_KERNEL);
7700         if (!q_vector)
7701                 return -ENOMEM;
7702
7703         q_vector->vsi = vsi;
7704         q_vector->v_idx = v_idx;
7705         cpumask_set_cpu(v_idx, &q_vector->affinity_mask);
7706         if (vsi->netdev)
7707                 netif_napi_add(vsi->netdev, &q_vector->napi,
7708                                i40e_napi_poll, NAPI_POLL_WEIGHT);
7709
7710         q_vector->rx.latency_range = I40E_LOW_LATENCY;
7711         q_vector->tx.latency_range = I40E_LOW_LATENCY;
7712
7713         /* tie q_vector and vsi together */
7714         vsi->q_vectors[v_idx] = q_vector;
7715
7716         return 0;
7717 }
7718
7719 /**
7720  * i40e_vsi_alloc_q_vectors - Allocate memory for interrupt vectors
7721  * @vsi: the VSI being configured
7722  *
7723  * We allocate one q_vector per queue interrupt.  If allocation fails we
7724  * return -ENOMEM.
7725  **/
7726 static int i40e_vsi_alloc_q_vectors(struct i40e_vsi *vsi)
7727 {
7728         struct i40e_pf *pf = vsi->back;
7729         int v_idx, num_q_vectors;
7730         int err;
7731
7732         /* if not MSIX, give the one vector only to the LAN VSI */
7733         if (pf->flags & I40E_FLAG_MSIX_ENABLED)
7734                 num_q_vectors = vsi->num_q_vectors;
7735         else if (vsi == pf->vsi[pf->lan_vsi])
7736                 num_q_vectors = 1;
7737         else
7738                 return -EINVAL;
7739
7740         for (v_idx = 0; v_idx < num_q_vectors; v_idx++) {
7741                 err = i40e_vsi_alloc_q_vector(vsi, v_idx);
7742                 if (err)
7743                         goto err_out;
7744         }
7745
7746         return 0;
7747
7748 err_out:
7749         while (v_idx--)
7750                 i40e_free_q_vector(vsi, v_idx);
7751
7752         return err;
7753 }
7754
7755 /**
7756  * i40e_init_interrupt_scheme - Determine proper interrupt scheme
7757  * @pf: board private structure to initialize
7758  **/
7759 static int i40e_init_interrupt_scheme(struct i40e_pf *pf)
7760 {
7761         int vectors = 0;
7762         ssize_t size;
7763
7764         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
7765                 vectors = i40e_init_msix(pf);
7766                 if (vectors < 0) {
7767                         pf->flags &= ~(I40E_FLAG_MSIX_ENABLED   |
7768 #ifdef I40E_FCOE
7769                                        I40E_FLAG_FCOE_ENABLED   |
7770 #endif
7771                                        I40E_FLAG_RSS_ENABLED    |
7772                                        I40E_FLAG_DCB_CAPABLE    |
7773                                        I40E_FLAG_SRIOV_ENABLED  |
7774                                        I40E_FLAG_FD_SB_ENABLED  |
7775                                        I40E_FLAG_FD_ATR_ENABLED |
7776                                        I40E_FLAG_VMDQ_ENABLED);
7777
7778                         /* rework the queue expectations without MSIX */
7779                         i40e_determine_queue_usage(pf);
7780                 }
7781         }
7782
7783         if (!(pf->flags & I40E_FLAG_MSIX_ENABLED) &&
7784             (pf->flags & I40E_FLAG_MSI_ENABLED)) {
7785                 dev_info(&pf->pdev->dev, "MSI-X not available, trying MSI\n");
7786                 vectors = pci_enable_msi(pf->pdev);
7787                 if (vectors < 0) {
7788                         dev_info(&pf->pdev->dev, "MSI init failed - %d\n",
7789                                  vectors);
7790                         pf->flags &= ~I40E_FLAG_MSI_ENABLED;
7791                 }
7792                 vectors = 1;  /* one MSI or Legacy vector */
7793         }
7794
7795         if (!(pf->flags & (I40E_FLAG_MSIX_ENABLED | I40E_FLAG_MSI_ENABLED)))
7796                 dev_info(&pf->pdev->dev, "MSI-X and MSI not available, falling back to Legacy IRQ\n");
7797
7798         /* set up vector assignment tracking */
7799         size = sizeof(struct i40e_lump_tracking) + (sizeof(u16) * vectors);
7800         pf->irq_pile = kzalloc(size, GFP_KERNEL);
7801         if (!pf->irq_pile) {
7802                 dev_err(&pf->pdev->dev, "error allocating irq_pile memory\n");
7803                 return -ENOMEM;
7804         }
7805         pf->irq_pile->num_entries = vectors;
7806         pf->irq_pile->search_hint = 0;
7807
7808         /* track first vector for misc interrupts, ignore return */
7809         (void)i40e_get_lump(pf, pf->irq_pile, 1, I40E_PILE_VALID_BIT - 1);
7810
7811         return 0;
7812 }
7813
7814 /**
7815  * i40e_setup_misc_vector - Setup the misc vector to handle non queue events
7816  * @pf: board private structure
7817  *
7818  * This sets up the handler for MSIX 0, which is used to manage the
7819  * non-queue interrupts, e.g. AdminQ and errors.  This is not used
7820  * when in MSI or Legacy interrupt mode.
7821  **/
7822 static int i40e_setup_misc_vector(struct i40e_pf *pf)
7823 {
7824         struct i40e_hw *hw = &pf->hw;
7825         int err = 0;
7826
7827         /* Only request the irq if this is the first time through, and
7828          * not when we're rebuilding after a Reset
7829          */
7830         if (!test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state)) {
7831                 err = request_irq(pf->msix_entries[0].vector,
7832                                   i40e_intr, 0, pf->int_name, pf);
7833                 if (err) {
7834                         dev_info(&pf->pdev->dev,
7835                                  "request_irq for %s failed: %d\n",
7836                                  pf->int_name, err);
7837                         return -EFAULT;
7838                 }
7839         }
7840
7841         i40e_enable_misc_int_causes(pf);
7842
7843         /* associate no queues to the misc vector */
7844         wr32(hw, I40E_PFINT_LNKLST0, I40E_QUEUE_END_OF_LIST);
7845         wr32(hw, I40E_PFINT_ITR0(I40E_RX_ITR), I40E_ITR_8K);
7846
7847         i40e_flush(hw);
7848
7849         i40e_irq_dynamic_enable_icr0(pf);
7850
7851         return err;
7852 }
7853
7854 /**
7855  * i40e_config_rss_aq - Prepare for RSS using AQ commands
7856  * @vsi: vsi structure
7857  * @seed: RSS hash seed
7858  **/
7859 static int i40e_config_rss_aq(struct i40e_vsi *vsi, const u8 *seed,
7860                               u8 *lut, u16 lut_size)
7861 {
7862         struct i40e_aqc_get_set_rss_key_data rss_key;
7863         struct i40e_pf *pf = vsi->back;
7864         struct i40e_hw *hw = &pf->hw;
7865         bool pf_lut = false;
7866         u8 *rss_lut;
7867         int ret, i;
7868
7869         memset(&rss_key, 0, sizeof(rss_key));
7870         memcpy(&rss_key, seed, sizeof(rss_key));
7871
7872         rss_lut = kzalloc(pf->rss_table_size, GFP_KERNEL);
7873         if (!rss_lut)
7874                 return -ENOMEM;
7875
7876         /* Populate the LUT with max no. of queues in round robin fashion */
7877         for (i = 0; i < vsi->rss_table_size; i++)
7878                 rss_lut[i] = i % vsi->rss_size;
7879
7880         ret = i40e_aq_set_rss_key(hw, vsi->id, &rss_key);
7881         if (ret) {
7882                 dev_info(&pf->pdev->dev,
7883                          "Cannot set RSS key, err %s aq_err %s\n",
7884                          i40e_stat_str(&pf->hw, ret),
7885                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
7886                 goto config_rss_aq_out;
7887         }
7888
7889         if (vsi->type == I40E_VSI_MAIN)
7890                 pf_lut = true;
7891
7892         ret = i40e_aq_set_rss_lut(hw, vsi->id, pf_lut, rss_lut,
7893                                   vsi->rss_table_size);
7894         if (ret)
7895                 dev_info(&pf->pdev->dev,
7896                          "Cannot set RSS lut, err %s aq_err %s\n",
7897                          i40e_stat_str(&pf->hw, ret),
7898                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
7899
7900 config_rss_aq_out:
7901         kfree(rss_lut);
7902         return ret;
7903 }
7904
7905 /**
7906  * i40e_vsi_config_rss - Prepare for VSI(VMDq) RSS if used
7907  * @vsi: VSI structure
7908  **/
7909 static int i40e_vsi_config_rss(struct i40e_vsi *vsi)
7910 {
7911         u8 seed[I40E_HKEY_ARRAY_SIZE];
7912         struct i40e_pf *pf = vsi->back;
7913         u8 *lut;
7914         int ret;
7915
7916         if (!(pf->flags & I40E_FLAG_RSS_AQ_CAPABLE))
7917                 return 0;
7918
7919         lut = kzalloc(vsi->rss_table_size, GFP_KERNEL);
7920         if (!lut)
7921                 return -ENOMEM;
7922
7923         i40e_fill_rss_lut(pf, lut, vsi->rss_table_size, vsi->rss_size);
7924         netdev_rss_key_fill((void *)seed, I40E_HKEY_ARRAY_SIZE);
7925         vsi->rss_size = min_t(int, pf->alloc_rss_size, vsi->num_queue_pairs);
7926         ret = i40e_config_rss_aq(vsi, seed, lut, vsi->rss_table_size);
7927         kfree(lut);
7928
7929         return ret;
7930 }
7931
7932 /**
7933  * i40e_config_rss_reg - Configure RSS keys and lut by writing registers
7934  * @vsi: Pointer to vsi structure
7935  * @seed: RSS hash seed
7936  * @lut: Lookup table
7937  * @lut_size: Lookup table size
7938  *
7939  * Returns 0 on success, negative on failure
7940  **/
7941 static int i40e_config_rss_reg(struct i40e_vsi *vsi, const u8 *seed,
7942                                const u8 *lut, u16 lut_size)
7943 {
7944         struct i40e_pf *pf = vsi->back;
7945         struct i40e_hw *hw = &pf->hw;
7946         u8 i;
7947
7948         /* Fill out hash function seed */
7949         if (seed) {
7950                 u32 *seed_dw = (u32 *)seed;
7951
7952                 for (i = 0; i <= I40E_PFQF_HKEY_MAX_INDEX; i++)
7953                         wr32(hw, I40E_PFQF_HKEY(i), seed_dw[i]);
7954         }
7955
7956         if (lut) {
7957                 u32 *lut_dw = (u32 *)lut;
7958
7959                 if (lut_size != I40E_HLUT_ARRAY_SIZE)
7960                         return -EINVAL;
7961
7962                 for (i = 0; i <= I40E_PFQF_HLUT_MAX_INDEX; i++)
7963                         wr32(hw, I40E_PFQF_HLUT(i), lut_dw[i]);
7964         }
7965         i40e_flush(hw);
7966
7967         return 0;
7968 }
7969
7970 /**
7971  * i40e_get_rss_reg - Get the RSS keys and lut by reading registers
7972  * @vsi: Pointer to VSI structure
7973  * @seed: Buffer to store the keys
7974  * @lut: Buffer to store the lookup table entries
7975  * @lut_size: Size of buffer to store the lookup table entries
7976  *
7977  * Returns 0 on success, negative on failure
7978  */
7979 static int i40e_get_rss_reg(struct i40e_vsi *vsi, u8 *seed,
7980                             u8 *lut, u16 lut_size)
7981 {
7982         struct i40e_pf *pf = vsi->back;
7983         struct i40e_hw *hw = &pf->hw;
7984         u16 i;
7985
7986         if (seed) {
7987                 u32 *seed_dw = (u32 *)seed;
7988
7989                 for (i = 0; i <= I40E_PFQF_HKEY_MAX_INDEX; i++)
7990                         seed_dw[i] = rd32(hw, I40E_PFQF_HKEY(i));
7991         }
7992         if (lut) {
7993                 u32 *lut_dw = (u32 *)lut;
7994
7995                 if (lut_size != I40E_HLUT_ARRAY_SIZE)
7996                         return -EINVAL;
7997                 for (i = 0; i <= I40E_PFQF_HLUT_MAX_INDEX; i++)
7998                         lut_dw[i] = rd32(hw, I40E_PFQF_HLUT(i));
7999         }
8000
8001         return 0;
8002 }
8003
8004 /**
8005  * i40e_config_rss - Configure RSS keys and lut
8006  * @vsi: Pointer to VSI structure
8007  * @seed: RSS hash seed
8008  * @lut: Lookup table
8009  * @lut_size: Lookup table size
8010  *
8011  * Returns 0 on success, negative on failure
8012  */
8013 int i40e_config_rss(struct i40e_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size)
8014 {
8015         struct i40e_pf *pf = vsi->back;
8016
8017         if (pf->flags & I40E_FLAG_RSS_AQ_CAPABLE)
8018                 return i40e_config_rss_aq(vsi, seed, lut, lut_size);
8019         else
8020                 return i40e_config_rss_reg(vsi, seed, lut, lut_size);
8021 }
8022
8023 /**
8024  * i40e_get_rss - Get RSS keys and lut
8025  * @vsi: Pointer to VSI structure
8026  * @seed: Buffer to store the keys
8027  * @lut: Buffer to store the lookup table entries
8028  * lut_size: Size of buffer to store the lookup table entries
8029  *
8030  * Returns 0 on success, negative on failure
8031  */
8032 int i40e_get_rss(struct i40e_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size)
8033 {
8034         return i40e_get_rss_reg(vsi, seed, lut, lut_size);
8035 }
8036
8037 /**
8038  * i40e_fill_rss_lut - Fill the RSS lookup table with default values
8039  * @pf: Pointer to board private structure
8040  * @lut: Lookup table
8041  * @rss_table_size: Lookup table size
8042  * @rss_size: Range of queue number for hashing
8043  */
8044 static void i40e_fill_rss_lut(struct i40e_pf *pf, u8 *lut,
8045                               u16 rss_table_size, u16 rss_size)
8046 {
8047         u16 i;
8048
8049         for (i = 0; i < rss_table_size; i++)
8050                 lut[i] = i % rss_size;
8051 }
8052
8053 /**
8054  * i40e_pf_config_rss - Prepare for RSS if used
8055  * @pf: board private structure
8056  **/
8057 static int i40e_pf_config_rss(struct i40e_pf *pf)
8058 {
8059         struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
8060         u8 seed[I40E_HKEY_ARRAY_SIZE];
8061         u8 *lut;
8062         struct i40e_hw *hw = &pf->hw;
8063         u32 reg_val;
8064         u64 hena;
8065         int ret;
8066
8067         /* By default we enable TCP/UDP with IPv4/IPv6 ptypes */
8068         hena = (u64)rd32(hw, I40E_PFQF_HENA(0)) |
8069                 ((u64)rd32(hw, I40E_PFQF_HENA(1)) << 32);
8070         hena |= i40e_pf_get_default_rss_hena(pf);
8071
8072         wr32(hw, I40E_PFQF_HENA(0), (u32)hena);
8073         wr32(hw, I40E_PFQF_HENA(1), (u32)(hena >> 32));
8074
8075         /* Determine the RSS table size based on the hardware capabilities */
8076         reg_val = rd32(hw, I40E_PFQF_CTL_0);
8077         reg_val = (pf->rss_table_size == 512) ?
8078                         (reg_val | I40E_PFQF_CTL_0_HASHLUTSIZE_512) :
8079                         (reg_val & ~I40E_PFQF_CTL_0_HASHLUTSIZE_512);
8080         wr32(hw, I40E_PFQF_CTL_0, reg_val);
8081
8082         /* Determine the RSS size of the VSI */
8083         if (!vsi->rss_size)
8084                 vsi->rss_size = min_t(int, pf->alloc_rss_size,
8085                                       vsi->num_queue_pairs);
8086
8087         lut = kzalloc(vsi->rss_table_size, GFP_KERNEL);
8088         if (!lut)
8089                 return -ENOMEM;
8090
8091         /* Use user configured lut if there is one, otherwise use default */
8092         if (vsi->rss_lut_user)
8093                 memcpy(lut, vsi->rss_lut_user, vsi->rss_table_size);
8094         else
8095                 i40e_fill_rss_lut(pf, lut, vsi->rss_table_size, vsi->rss_size);
8096
8097         /* Use user configured hash key if there is one, otherwise
8098          * use default.
8099          */
8100         if (vsi->rss_hkey_user)
8101                 memcpy(seed, vsi->rss_hkey_user, I40E_HKEY_ARRAY_SIZE);
8102         else
8103                 netdev_rss_key_fill((void *)seed, I40E_HKEY_ARRAY_SIZE);
8104         ret = i40e_config_rss(vsi, seed, lut, vsi->rss_table_size);
8105         kfree(lut);
8106
8107         return ret;
8108 }
8109
8110 /**
8111  * i40e_reconfig_rss_queues - change number of queues for rss and rebuild
8112  * @pf: board private structure
8113  * @queue_count: the requested queue count for rss.
8114  *
8115  * returns 0 if rss is not enabled, if enabled returns the final rss queue
8116  * count which may be different from the requested queue count.
8117  **/
8118 int i40e_reconfig_rss_queues(struct i40e_pf *pf, int queue_count)
8119 {
8120         struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
8121         int new_rss_size;
8122
8123         if (!(pf->flags & I40E_FLAG_RSS_ENABLED))
8124                 return 0;
8125
8126         new_rss_size = min_t(int, queue_count, pf->rss_size_max);
8127
8128         if (queue_count != vsi->num_queue_pairs) {
8129                 vsi->req_queue_pairs = queue_count;
8130                 i40e_prep_for_reset(pf);
8131
8132                 pf->alloc_rss_size = new_rss_size;
8133
8134                 i40e_reset_and_rebuild(pf, true);
8135
8136                 /* Discard the user configured hash keys and lut, if less
8137                  * queues are enabled.
8138                  */
8139                 if (queue_count < vsi->rss_size) {
8140                         i40e_clear_rss_config_user(vsi);
8141                         dev_dbg(&pf->pdev->dev,
8142                                 "discard user configured hash keys and lut\n");
8143                 }
8144
8145                 /* Reset vsi->rss_size, as number of enabled queues changed */
8146                 vsi->rss_size = min_t(int, pf->alloc_rss_size,
8147                                       vsi->num_queue_pairs);
8148
8149                 i40e_pf_config_rss(pf);
8150         }
8151         dev_info(&pf->pdev->dev, "RSS count/HW max RSS count:  %d/%d\n",
8152                  pf->alloc_rss_size, pf->rss_size_max);
8153         return pf->alloc_rss_size;
8154 }
8155
8156 /**
8157  * i40e_get_npar_bw_setting - Retrieve BW settings for this PF partition
8158  * @pf: board private structure
8159  **/
8160 i40e_status i40e_get_npar_bw_setting(struct i40e_pf *pf)
8161 {
8162         i40e_status status;
8163         bool min_valid, max_valid;
8164         u32 max_bw, min_bw;
8165
8166         status = i40e_read_bw_from_alt_ram(&pf->hw, &max_bw, &min_bw,
8167                                            &min_valid, &max_valid);
8168
8169         if (!status) {
8170                 if (min_valid)
8171                         pf->npar_min_bw = min_bw;
8172                 if (max_valid)
8173                         pf->npar_max_bw = max_bw;
8174         }
8175
8176         return status;
8177 }
8178
8179 /**
8180  * i40e_set_npar_bw_setting - Set BW settings for this PF partition
8181  * @pf: board private structure
8182  **/
8183 i40e_status i40e_set_npar_bw_setting(struct i40e_pf *pf)
8184 {
8185         struct i40e_aqc_configure_partition_bw_data bw_data;
8186         i40e_status status;
8187
8188         /* Set the valid bit for this PF */
8189         bw_data.pf_valid_bits = cpu_to_le16(BIT(pf->hw.pf_id));
8190         bw_data.max_bw[pf->hw.pf_id] = pf->npar_max_bw & I40E_ALT_BW_VALUE_MASK;
8191         bw_data.min_bw[pf->hw.pf_id] = pf->npar_min_bw & I40E_ALT_BW_VALUE_MASK;
8192
8193         /* Set the new bandwidths */
8194         status = i40e_aq_configure_partition_bw(&pf->hw, &bw_data, NULL);
8195
8196         return status;
8197 }
8198
8199 /**
8200  * i40e_commit_npar_bw_setting - Commit BW settings for this PF partition
8201  * @pf: board private structure
8202  **/
8203 i40e_status i40e_commit_npar_bw_setting(struct i40e_pf *pf)
8204 {
8205         /* Commit temporary BW setting to permanent NVM image */
8206         enum i40e_admin_queue_err last_aq_status;
8207         i40e_status ret;
8208         u16 nvm_word;
8209
8210         if (pf->hw.partition_id != 1) {
8211                 dev_info(&pf->pdev->dev,
8212                          "Commit BW only works on partition 1! This is partition %d",
8213                          pf->hw.partition_id);
8214                 ret = I40E_NOT_SUPPORTED;
8215                 goto bw_commit_out;
8216         }
8217
8218         /* Acquire NVM for read access */
8219         ret = i40e_acquire_nvm(&pf->hw, I40E_RESOURCE_READ);
8220         last_aq_status = pf->hw.aq.asq_last_status;
8221         if (ret) {
8222                 dev_info(&pf->pdev->dev,
8223                          "Cannot acquire NVM for read access, err %s aq_err %s\n",
8224                          i40e_stat_str(&pf->hw, ret),
8225                          i40e_aq_str(&pf->hw, last_aq_status));
8226                 goto bw_commit_out;
8227         }
8228
8229         /* Read word 0x10 of NVM - SW compatibility word 1 */
8230         ret = i40e_aq_read_nvm(&pf->hw,
8231                                I40E_SR_NVM_CONTROL_WORD,
8232                                0x10, sizeof(nvm_word), &nvm_word,
8233                                false, NULL);
8234         /* Save off last admin queue command status before releasing
8235          * the NVM
8236          */
8237         last_aq_status = pf->hw.aq.asq_last_status;
8238         i40e_release_nvm(&pf->hw);
8239         if (ret) {
8240                 dev_info(&pf->pdev->dev, "NVM read error, err %s aq_err %s\n",
8241                          i40e_stat_str(&pf->hw, ret),
8242                          i40e_aq_str(&pf->hw, last_aq_status));
8243                 goto bw_commit_out;
8244         }
8245
8246         /* Wait a bit for NVM release to complete */
8247         msleep(50);
8248
8249         /* Acquire NVM for write access */
8250         ret = i40e_acquire_nvm(&pf->hw, I40E_RESOURCE_WRITE);
8251         last_aq_status = pf->hw.aq.asq_last_status;
8252         if (ret) {
8253                 dev_info(&pf->pdev->dev,
8254                          "Cannot acquire NVM for write access, err %s aq_err %s\n",
8255                          i40e_stat_str(&pf->hw, ret),
8256                          i40e_aq_str(&pf->hw, last_aq_status));
8257                 goto bw_commit_out;
8258         }
8259         /* Write it back out unchanged to initiate update NVM,
8260          * which will force a write of the shadow (alt) RAM to
8261          * the NVM - thus storing the bandwidth values permanently.
8262          */
8263         ret = i40e_aq_update_nvm(&pf->hw,
8264                                  I40E_SR_NVM_CONTROL_WORD,
8265                                  0x10, sizeof(nvm_word),
8266                                  &nvm_word, true, NULL);
8267         /* Save off last admin queue command status before releasing
8268          * the NVM
8269          */
8270         last_aq_status = pf->hw.aq.asq_last_status;
8271         i40e_release_nvm(&pf->hw);
8272         if (ret)
8273                 dev_info(&pf->pdev->dev,
8274                          "BW settings NOT SAVED, err %s aq_err %s\n",
8275                          i40e_stat_str(&pf->hw, ret),
8276                          i40e_aq_str(&pf->hw, last_aq_status));
8277 bw_commit_out:
8278
8279         return ret;
8280 }
8281
8282 /**
8283  * i40e_sw_init - Initialize general software structures (struct i40e_pf)
8284  * @pf: board private structure to initialize
8285  *
8286  * i40e_sw_init initializes the Adapter private data structure.
8287  * Fields are initialized based on PCI device information and
8288  * OS network device settings (MTU size).
8289  **/
8290 static int i40e_sw_init(struct i40e_pf *pf)
8291 {
8292         int err = 0;
8293         int size;
8294
8295         pf->msg_enable = netif_msg_init(I40E_DEFAULT_MSG_ENABLE,
8296                                 (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK));
8297         pf->hw.debug_mask = pf->msg_enable | I40E_DEBUG_DIAG;
8298         if (debug != -1 && debug != I40E_DEFAULT_MSG_ENABLE) {
8299                 if (I40E_DEBUG_USER & debug)
8300                         pf->hw.debug_mask = debug;
8301                 pf->msg_enable = netif_msg_init((debug & ~I40E_DEBUG_USER),
8302                                                 I40E_DEFAULT_MSG_ENABLE);
8303         }
8304
8305         /* Set default capability flags */
8306         pf->flags = I40E_FLAG_RX_CSUM_ENABLED |
8307                     I40E_FLAG_MSI_ENABLED     |
8308                     I40E_FLAG_LINK_POLLING_ENABLED |
8309                     I40E_FLAG_MSIX_ENABLED;
8310
8311         if (iommu_present(&pci_bus_type))
8312                 pf->flags |= I40E_FLAG_RX_PS_ENABLED;
8313         else
8314                 pf->flags |= I40E_FLAG_RX_1BUF_ENABLED;
8315
8316         /* Set default ITR */
8317         pf->rx_itr_default = I40E_ITR_DYNAMIC | I40E_ITR_RX_DEF;
8318         pf->tx_itr_default = I40E_ITR_DYNAMIC | I40E_ITR_TX_DEF;
8319
8320         /* Depending on PF configurations, it is possible that the RSS
8321          * maximum might end up larger than the available queues
8322          */
8323         pf->rss_size_max = BIT(pf->hw.func_caps.rss_table_entry_width);
8324         pf->alloc_rss_size = 1;
8325         pf->rss_table_size = pf->hw.func_caps.rss_table_size;
8326         pf->rss_size_max = min_t(int, pf->rss_size_max,
8327                                  pf->hw.func_caps.num_tx_qp);
8328         if (pf->hw.func_caps.rss) {
8329                 pf->flags |= I40E_FLAG_RSS_ENABLED;
8330                 pf->alloc_rss_size = min_t(int, pf->rss_size_max,
8331                                            num_online_cpus());
8332         }
8333
8334         /* MFP mode enabled */
8335         if (pf->hw.func_caps.npar_enable || pf->hw.func_caps.flex10_enable) {
8336                 pf->flags |= I40E_FLAG_MFP_ENABLED;
8337                 dev_info(&pf->pdev->dev, "MFP mode Enabled\n");
8338                 if (i40e_get_npar_bw_setting(pf))
8339                         dev_warn(&pf->pdev->dev,
8340                                  "Could not get NPAR bw settings\n");
8341                 else
8342                         dev_info(&pf->pdev->dev,
8343                                  "Min BW = %8.8x, Max BW = %8.8x\n",
8344                                  pf->npar_min_bw, pf->npar_max_bw);
8345         }
8346
8347         /* FW/NVM is not yet fixed in this regard */
8348         if ((pf->hw.func_caps.fd_filters_guaranteed > 0) ||
8349             (pf->hw.func_caps.fd_filters_best_effort > 0)) {
8350                 pf->flags |= I40E_FLAG_FD_ATR_ENABLED;
8351                 pf->atr_sample_rate = I40E_DEFAULT_ATR_SAMPLE_RATE;
8352                 if (pf->flags & I40E_FLAG_MFP_ENABLED &&
8353                     pf->hw.num_partitions > 1)
8354                         dev_info(&pf->pdev->dev,
8355                                  "Flow Director Sideband mode Disabled in MFP mode\n");
8356                 else
8357                         pf->flags |= I40E_FLAG_FD_SB_ENABLED;
8358                 pf->fdir_pf_filter_count =
8359                                  pf->hw.func_caps.fd_filters_guaranteed;
8360                 pf->hw.fdir_shared_filter_count =
8361                                  pf->hw.func_caps.fd_filters_best_effort;
8362         }
8363
8364         if (pf->hw.func_caps.vmdq) {
8365                 pf->num_vmdq_vsis = I40E_DEFAULT_NUM_VMDQ_VSI;
8366                 pf->flags |= I40E_FLAG_VMDQ_ENABLED;
8367                 pf->num_vmdq_qps = i40e_default_queues_per_vmdq(pf);
8368         }
8369
8370 #ifdef I40E_FCOE
8371         i40e_init_pf_fcoe(pf);
8372
8373 #endif /* I40E_FCOE */
8374 #ifdef CONFIG_PCI_IOV
8375         if (pf->hw.func_caps.num_vfs && pf->hw.partition_id == 1) {
8376                 pf->num_vf_qps = I40E_DEFAULT_QUEUES_PER_VF;
8377                 pf->flags |= I40E_FLAG_SRIOV_ENABLED;
8378                 pf->num_req_vfs = min_t(int,
8379                                         pf->hw.func_caps.num_vfs,
8380                                         I40E_MAX_VF_COUNT);
8381         }
8382 #endif /* CONFIG_PCI_IOV */
8383         if (pf->hw.mac.type == I40E_MAC_X722) {
8384                 pf->flags |= I40E_FLAG_RSS_AQ_CAPABLE |
8385                              I40E_FLAG_128_QP_RSS_CAPABLE |
8386                              I40E_FLAG_HW_ATR_EVICT_CAPABLE |
8387                              I40E_FLAG_OUTER_UDP_CSUM_CAPABLE |
8388                              I40E_FLAG_WB_ON_ITR_CAPABLE |
8389                              I40E_FLAG_MULTIPLE_TCP_UDP_RSS_PCTYPE |
8390                              I40E_FLAG_GENEVE_OFFLOAD_CAPABLE;
8391         }
8392         pf->eeprom_version = 0xDEAD;
8393         pf->lan_veb = I40E_NO_VEB;
8394         pf->lan_vsi = I40E_NO_VSI;
8395
8396         /* By default FW has this off for performance reasons */
8397         pf->flags &= ~I40E_FLAG_VEB_STATS_ENABLED;
8398
8399         /* set up queue assignment tracking */
8400         size = sizeof(struct i40e_lump_tracking)
8401                 + (sizeof(u16) * pf->hw.func_caps.num_tx_qp);
8402         pf->qp_pile = kzalloc(size, GFP_KERNEL);
8403         if (!pf->qp_pile) {
8404                 err = -ENOMEM;
8405                 goto sw_init_done;
8406         }
8407         pf->qp_pile->num_entries = pf->hw.func_caps.num_tx_qp;
8408         pf->qp_pile->search_hint = 0;
8409
8410         pf->tx_timeout_recovery_level = 1;
8411
8412         mutex_init(&pf->switch_mutex);
8413
8414         /* If NPAR is enabled nudge the Tx scheduler */
8415         if (pf->hw.func_caps.npar_enable && (!i40e_get_npar_bw_setting(pf)))
8416                 i40e_set_npar_bw_setting(pf);
8417
8418 sw_init_done:
8419         return err;
8420 }
8421
8422 /**
8423  * i40e_set_ntuple - set the ntuple feature flag and take action
8424  * @pf: board private structure to initialize
8425  * @features: the feature set that the stack is suggesting
8426  *
8427  * returns a bool to indicate if reset needs to happen
8428  **/
8429 bool i40e_set_ntuple(struct i40e_pf *pf, netdev_features_t features)
8430 {
8431         bool need_reset = false;
8432
8433         /* Check if Flow Director n-tuple support was enabled or disabled.  If
8434          * the state changed, we need to reset.
8435          */
8436         if (features & NETIF_F_NTUPLE) {
8437                 /* Enable filters and mark for reset */
8438                 if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
8439                         need_reset = true;
8440                 pf->flags |= I40E_FLAG_FD_SB_ENABLED;
8441         } else {
8442                 /* turn off filters, mark for reset and clear SW filter list */
8443                 if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
8444                         need_reset = true;
8445                         i40e_fdir_filter_exit(pf);
8446                 }
8447                 pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
8448                 pf->auto_disable_flags &= ~I40E_FLAG_FD_SB_ENABLED;
8449                 /* reset fd counters */
8450                 pf->fd_add_err = pf->fd_atr_cnt = pf->fd_tcp_rule = 0;
8451                 pf->fdir_pf_active_filters = 0;
8452                 pf->flags |= I40E_FLAG_FD_ATR_ENABLED;
8453                 if (I40E_DEBUG_FD & pf->hw.debug_mask)
8454                         dev_info(&pf->pdev->dev, "ATR re-enabled.\n");
8455                 /* if ATR was auto disabled it can be re-enabled. */
8456                 if ((pf->flags & I40E_FLAG_FD_ATR_ENABLED) &&
8457                     (pf->auto_disable_flags & I40E_FLAG_FD_ATR_ENABLED))
8458                         pf->auto_disable_flags &= ~I40E_FLAG_FD_ATR_ENABLED;
8459         }
8460         return need_reset;
8461 }
8462
8463 /**
8464  * i40e_set_features - set the netdev feature flags
8465  * @netdev: ptr to the netdev being adjusted
8466  * @features: the feature set that the stack is suggesting
8467  **/
8468 static int i40e_set_features(struct net_device *netdev,
8469                              netdev_features_t features)
8470 {
8471         struct i40e_netdev_priv *np = netdev_priv(netdev);
8472         struct i40e_vsi *vsi = np->vsi;
8473         struct i40e_pf *pf = vsi->back;
8474         bool need_reset;
8475
8476         if (features & NETIF_F_HW_VLAN_CTAG_RX)
8477                 i40e_vlan_stripping_enable(vsi);
8478         else
8479                 i40e_vlan_stripping_disable(vsi);
8480
8481         need_reset = i40e_set_ntuple(pf, features);
8482
8483         if (need_reset)
8484                 i40e_do_reset(pf, BIT_ULL(__I40E_PF_RESET_REQUESTED));
8485
8486         return 0;
8487 }
8488
8489 #if IS_ENABLED(CONFIG_VXLAN) || IS_ENABLED(CONFIG_GENEVE)
8490 /**
8491  * i40e_get_udp_port_idx - Lookup a possibly offloaded for Rx UDP port
8492  * @pf: board private structure
8493  * @port: The UDP port to look up
8494  *
8495  * Returns the index number or I40E_MAX_PF_UDP_OFFLOAD_PORTS if port not found
8496  **/
8497 static u8 i40e_get_udp_port_idx(struct i40e_pf *pf, __be16 port)
8498 {
8499         u8 i;
8500
8501         for (i = 0; i < I40E_MAX_PF_UDP_OFFLOAD_PORTS; i++) {
8502                 if (pf->udp_ports[i].index == port)
8503                         return i;
8504         }
8505
8506         return i;
8507 }
8508
8509 #endif
8510 /**
8511  * i40e_add_vxlan_port - Get notifications about VXLAN ports that come up
8512  * @netdev: This physical port's netdev
8513  * @sa_family: Socket Family that VXLAN is notifying us about
8514  * @port: New UDP port number that VXLAN started listening to
8515  **/
8516 static void i40e_add_vxlan_port(struct net_device *netdev,
8517                                 sa_family_t sa_family, __be16 port)
8518 {
8519 #if IS_ENABLED(CONFIG_VXLAN)
8520         struct i40e_netdev_priv *np = netdev_priv(netdev);
8521         struct i40e_vsi *vsi = np->vsi;
8522         struct i40e_pf *pf = vsi->back;
8523         u8 next_idx;
8524         u8 idx;
8525
8526         if (sa_family == AF_INET6)
8527                 return;
8528
8529         idx = i40e_get_udp_port_idx(pf, port);
8530
8531         /* Check if port already exists */
8532         if (idx < I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
8533                 netdev_info(netdev, "vxlan port %d already offloaded\n",
8534                             ntohs(port));
8535                 return;
8536         }
8537
8538         /* Now check if there is space to add the new port */
8539         next_idx = i40e_get_udp_port_idx(pf, 0);
8540
8541         if (next_idx == I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
8542                 netdev_info(netdev, "maximum number of vxlan UDP ports reached, not adding port %d\n",
8543                             ntohs(port));
8544                 return;
8545         }
8546
8547         /* New port: add it and mark its index in the bitmap */
8548         pf->udp_ports[next_idx].index = port;
8549         pf->udp_ports[next_idx].type = I40E_AQC_TUNNEL_TYPE_VXLAN;
8550         pf->pending_udp_bitmap |= BIT_ULL(next_idx);
8551         pf->flags |= I40E_FLAG_UDP_FILTER_SYNC;
8552 #endif
8553 }
8554
8555 /**
8556  * i40e_del_vxlan_port - Get notifications about VXLAN ports that go away
8557  * @netdev: This physical port's netdev
8558  * @sa_family: Socket Family that VXLAN is notifying us about
8559  * @port: UDP port number that VXLAN stopped listening to
8560  **/
8561 static void i40e_del_vxlan_port(struct net_device *netdev,
8562                                 sa_family_t sa_family, __be16 port)
8563 {
8564 #if IS_ENABLED(CONFIG_VXLAN)
8565         struct i40e_netdev_priv *np = netdev_priv(netdev);
8566         struct i40e_vsi *vsi = np->vsi;
8567         struct i40e_pf *pf = vsi->back;
8568         u8 idx;
8569
8570         if (sa_family == AF_INET6)
8571                 return;
8572
8573         idx = i40e_get_udp_port_idx(pf, port);
8574
8575         /* Check if port already exists */
8576         if (idx < I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
8577                 /* if port exists, set it to 0 (mark for deletion)
8578                  * and make it pending
8579                  */
8580                 pf->udp_ports[idx].index = 0;
8581                 pf->pending_udp_bitmap |= BIT_ULL(idx);
8582                 pf->flags |= I40E_FLAG_UDP_FILTER_SYNC;
8583         } else {
8584                 netdev_warn(netdev, "vxlan port %d was not found, not deleting\n",
8585                             ntohs(port));
8586         }
8587 #endif
8588 }
8589
8590 /**
8591  * i40e_add_geneve_port - Get notifications about GENEVE ports that come up
8592  * @netdev: This physical port's netdev
8593  * @sa_family: Socket Family that GENEVE is notifying us about
8594  * @port: New UDP port number that GENEVE started listening to
8595  **/
8596 static void i40e_add_geneve_port(struct net_device *netdev,
8597                                  sa_family_t sa_family, __be16 port)
8598 {
8599 #if IS_ENABLED(CONFIG_GENEVE)
8600         struct i40e_netdev_priv *np = netdev_priv(netdev);
8601         struct i40e_vsi *vsi = np->vsi;
8602         struct i40e_pf *pf = vsi->back;
8603         u8 next_idx;
8604         u8 idx;
8605
8606         if (sa_family == AF_INET6)
8607                 return;
8608
8609         idx = i40e_get_udp_port_idx(pf, port);
8610
8611         /* Check if port already exists */
8612         if (idx < I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
8613                 netdev_info(netdev, "udp port %d already offloaded\n",
8614                             ntohs(port));
8615                 return;
8616         }
8617
8618         /* Now check if there is space to add the new port */
8619         next_idx = i40e_get_udp_port_idx(pf, 0);
8620
8621         if (next_idx == I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
8622                 netdev_info(netdev, "maximum number of UDP ports reached, not adding port %d\n",
8623                             ntohs(port));
8624                 return;
8625         }
8626
8627         /* New port: add it and mark its index in the bitmap */
8628         pf->udp_ports[next_idx].index = port;
8629         pf->udp_ports[next_idx].type = I40E_AQC_TUNNEL_TYPE_NGE;
8630         pf->pending_udp_bitmap |= BIT_ULL(next_idx);
8631         pf->flags |= I40E_FLAG_UDP_FILTER_SYNC;
8632
8633         dev_info(&pf->pdev->dev, "adding geneve port %d\n", ntohs(port));
8634 #endif
8635 }
8636
8637 /**
8638  * i40e_del_geneve_port - Get notifications about GENEVE ports that go away
8639  * @netdev: This physical port's netdev
8640  * @sa_family: Socket Family that GENEVE is notifying us about
8641  * @port: UDP port number that GENEVE stopped listening to
8642  **/
8643 static void i40e_del_geneve_port(struct net_device *netdev,
8644                                  sa_family_t sa_family, __be16 port)
8645 {
8646 #if IS_ENABLED(CONFIG_GENEVE)
8647         struct i40e_netdev_priv *np = netdev_priv(netdev);
8648         struct i40e_vsi *vsi = np->vsi;
8649         struct i40e_pf *pf = vsi->back;
8650         u8 idx;
8651
8652         if (sa_family == AF_INET6)
8653                 return;
8654
8655         idx = i40e_get_udp_port_idx(pf, port);
8656
8657         /* Check if port already exists */
8658         if (idx < I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
8659                 /* if port exists, set it to 0 (mark for deletion)
8660                  * and make it pending
8661                  */
8662                 pf->udp_ports[idx].index = 0;
8663                 pf->pending_udp_bitmap |= BIT_ULL(idx);
8664                 pf->flags |= I40E_FLAG_UDP_FILTER_SYNC;
8665
8666                 dev_info(&pf->pdev->dev, "deleting geneve port %d\n",
8667                          ntohs(port));
8668         } else {
8669                 netdev_warn(netdev, "geneve port %d was not found, not deleting\n",
8670                             ntohs(port));
8671         }
8672 #endif
8673 }
8674
8675 static int i40e_get_phys_port_id(struct net_device *netdev,
8676                                  struct netdev_phys_item_id *ppid)
8677 {
8678         struct i40e_netdev_priv *np = netdev_priv(netdev);
8679         struct i40e_pf *pf = np->vsi->back;
8680         struct i40e_hw *hw = &pf->hw;
8681
8682         if (!(pf->flags & I40E_FLAG_PORT_ID_VALID))
8683                 return -EOPNOTSUPP;
8684
8685         ppid->id_len = min_t(int, sizeof(hw->mac.port_addr), sizeof(ppid->id));
8686         memcpy(ppid->id, hw->mac.port_addr, ppid->id_len);
8687
8688         return 0;
8689 }
8690
8691 /**
8692  * i40e_ndo_fdb_add - add an entry to the hardware database
8693  * @ndm: the input from the stack
8694  * @tb: pointer to array of nladdr (unused)
8695  * @dev: the net device pointer
8696  * @addr: the MAC address entry being added
8697  * @flags: instructions from stack about fdb operation
8698  */
8699 static int i40e_ndo_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
8700                             struct net_device *dev,
8701                             const unsigned char *addr, u16 vid,
8702                             u16 flags)
8703 {
8704         struct i40e_netdev_priv *np = netdev_priv(dev);
8705         struct i40e_pf *pf = np->vsi->back;
8706         int err = 0;
8707
8708         if (!(pf->flags & I40E_FLAG_SRIOV_ENABLED))
8709                 return -EOPNOTSUPP;
8710
8711         if (vid) {
8712                 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
8713                 return -EINVAL;
8714         }
8715
8716         /* Hardware does not support aging addresses so if a
8717          * ndm_state is given only allow permanent addresses
8718          */
8719         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
8720                 netdev_info(dev, "FDB only supports static addresses\n");
8721                 return -EINVAL;
8722         }
8723
8724         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
8725                 err = dev_uc_add_excl(dev, addr);
8726         else if (is_multicast_ether_addr(addr))
8727                 err = dev_mc_add_excl(dev, addr);
8728         else
8729                 err = -EINVAL;
8730
8731         /* Only return duplicate errors if NLM_F_EXCL is set */
8732         if (err == -EEXIST && !(flags & NLM_F_EXCL))
8733                 err = 0;
8734
8735         return err;
8736 }
8737
8738 /**
8739  * i40e_ndo_bridge_setlink - Set the hardware bridge mode
8740  * @dev: the netdev being configured
8741  * @nlh: RTNL message
8742  *
8743  * Inserts a new hardware bridge if not already created and
8744  * enables the bridging mode requested (VEB or VEPA). If the
8745  * hardware bridge has already been inserted and the request
8746  * is to change the mode then that requires a PF reset to
8747  * allow rebuild of the components with required hardware
8748  * bridge mode enabled.
8749  **/
8750 static int i40e_ndo_bridge_setlink(struct net_device *dev,
8751                                    struct nlmsghdr *nlh,
8752                                    u16 flags)
8753 {
8754         struct i40e_netdev_priv *np = netdev_priv(dev);
8755         struct i40e_vsi *vsi = np->vsi;
8756         struct i40e_pf *pf = vsi->back;
8757         struct i40e_veb *veb = NULL;
8758         struct nlattr *attr, *br_spec;
8759         int i, rem;
8760
8761         /* Only for PF VSI for now */
8762         if (vsi->seid != pf->vsi[pf->lan_vsi]->seid)
8763                 return -EOPNOTSUPP;
8764
8765         /* Find the HW bridge for PF VSI */
8766         for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
8767                 if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
8768                         veb = pf->veb[i];
8769         }
8770
8771         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
8772
8773         nla_for_each_nested(attr, br_spec, rem) {
8774                 __u16 mode;
8775
8776                 if (nla_type(attr) != IFLA_BRIDGE_MODE)
8777                         continue;
8778
8779                 mode = nla_get_u16(attr);
8780                 if ((mode != BRIDGE_MODE_VEPA) &&
8781                     (mode != BRIDGE_MODE_VEB))
8782                         return -EINVAL;
8783
8784                 /* Insert a new HW bridge */
8785                 if (!veb) {
8786                         veb = i40e_veb_setup(pf, 0, vsi->uplink_seid, vsi->seid,
8787                                              vsi->tc_config.enabled_tc);
8788                         if (veb) {
8789                                 veb->bridge_mode = mode;
8790                                 i40e_config_bridge_mode(veb);
8791                         } else {
8792                                 /* No Bridge HW offload available */
8793                                 return -ENOENT;
8794                         }
8795                         break;
8796                 } else if (mode != veb->bridge_mode) {
8797                         /* Existing HW bridge but different mode needs reset */
8798                         veb->bridge_mode = mode;
8799                         /* TODO: If no VFs or VMDq VSIs, disallow VEB mode */
8800                         if (mode == BRIDGE_MODE_VEB)
8801                                 pf->flags |= I40E_FLAG_VEB_MODE_ENABLED;
8802                         else
8803                                 pf->flags &= ~I40E_FLAG_VEB_MODE_ENABLED;
8804                         i40e_do_reset(pf, BIT_ULL(__I40E_PF_RESET_REQUESTED));
8805                         break;
8806                 }
8807         }
8808
8809         return 0;
8810 }
8811
8812 /**
8813  * i40e_ndo_bridge_getlink - Get the hardware bridge mode
8814  * @skb: skb buff
8815  * @pid: process id
8816  * @seq: RTNL message seq #
8817  * @dev: the netdev being configured
8818  * @filter_mask: unused
8819  * @nlflags: netlink flags passed in
8820  *
8821  * Return the mode in which the hardware bridge is operating in
8822  * i.e VEB or VEPA.
8823  **/
8824 static int i40e_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
8825                                    struct net_device *dev,
8826                                    u32 __always_unused filter_mask,
8827                                    int nlflags)
8828 {
8829         struct i40e_netdev_priv *np = netdev_priv(dev);
8830         struct i40e_vsi *vsi = np->vsi;
8831         struct i40e_pf *pf = vsi->back;
8832         struct i40e_veb *veb = NULL;
8833         int i;
8834
8835         /* Only for PF VSI for now */
8836         if (vsi->seid != pf->vsi[pf->lan_vsi]->seid)
8837                 return -EOPNOTSUPP;
8838
8839         /* Find the HW bridge for the PF VSI */
8840         for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
8841                 if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
8842                         veb = pf->veb[i];
8843         }
8844
8845         if (!veb)
8846                 return 0;
8847
8848         return ndo_dflt_bridge_getlink(skb, pid, seq, dev, veb->bridge_mode,
8849                                        nlflags, 0, 0, filter_mask, NULL);
8850 }
8851
8852 /* Hardware supports L4 tunnel length of 128B (=2^7) which includes
8853  * inner mac plus all inner ethertypes.
8854  */
8855 #define I40E_MAX_TUNNEL_HDR_LEN 128
8856 /**
8857  * i40e_features_check - Validate encapsulated packet conforms to limits
8858  * @skb: skb buff
8859  * @dev: This physical port's netdev
8860  * @features: Offload features that the stack believes apply
8861  **/
8862 static netdev_features_t i40e_features_check(struct sk_buff *skb,
8863                                              struct net_device *dev,
8864                                              netdev_features_t features)
8865 {
8866         if (skb->encapsulation &&
8867             ((skb_inner_network_header(skb) - skb_transport_header(skb)) >
8868              I40E_MAX_TUNNEL_HDR_LEN))
8869                 return features & ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK);
8870
8871         return features;
8872 }
8873
8874 static const struct net_device_ops i40e_netdev_ops = {
8875         .ndo_open               = i40e_open,
8876         .ndo_stop               = i40e_close,
8877         .ndo_start_xmit         = i40e_lan_xmit_frame,
8878         .ndo_get_stats64        = i40e_get_netdev_stats_struct,
8879         .ndo_set_rx_mode        = i40e_set_rx_mode,
8880         .ndo_validate_addr      = eth_validate_addr,
8881         .ndo_set_mac_address    = i40e_set_mac,
8882         .ndo_change_mtu         = i40e_change_mtu,
8883         .ndo_do_ioctl           = i40e_ioctl,
8884         .ndo_tx_timeout         = i40e_tx_timeout,
8885         .ndo_vlan_rx_add_vid    = i40e_vlan_rx_add_vid,
8886         .ndo_vlan_rx_kill_vid   = i40e_vlan_rx_kill_vid,
8887 #ifdef CONFIG_NET_POLL_CONTROLLER
8888         .ndo_poll_controller    = i40e_netpoll,
8889 #endif
8890         .ndo_setup_tc           = i40e_setup_tc,
8891 #ifdef I40E_FCOE
8892         .ndo_fcoe_enable        = i40e_fcoe_enable,
8893         .ndo_fcoe_disable       = i40e_fcoe_disable,
8894 #endif
8895         .ndo_set_features       = i40e_set_features,
8896         .ndo_set_vf_mac         = i40e_ndo_set_vf_mac,
8897         .ndo_set_vf_vlan        = i40e_ndo_set_vf_port_vlan,
8898         .ndo_set_vf_rate        = i40e_ndo_set_vf_bw,
8899         .ndo_get_vf_config      = i40e_ndo_get_vf_config,
8900         .ndo_set_vf_link_state  = i40e_ndo_set_vf_link_state,
8901         .ndo_set_vf_spoofchk    = i40e_ndo_set_vf_spoofchk,
8902 #if IS_ENABLED(CONFIG_VXLAN)
8903         .ndo_add_vxlan_port     = i40e_add_vxlan_port,
8904         .ndo_del_vxlan_port     = i40e_del_vxlan_port,
8905 #endif
8906 #if IS_ENABLED(CONFIG_GENEVE)
8907         .ndo_add_geneve_port    = i40e_add_geneve_port,
8908         .ndo_del_geneve_port    = i40e_del_geneve_port,
8909 #endif
8910         .ndo_get_phys_port_id   = i40e_get_phys_port_id,
8911         .ndo_fdb_add            = i40e_ndo_fdb_add,
8912         .ndo_features_check     = i40e_features_check,
8913         .ndo_bridge_getlink     = i40e_ndo_bridge_getlink,
8914         .ndo_bridge_setlink     = i40e_ndo_bridge_setlink,
8915 };
8916
8917 /**
8918  * i40e_config_netdev - Setup the netdev flags
8919  * @vsi: the VSI being configured
8920  *
8921  * Returns 0 on success, negative value on failure
8922  **/
8923 static int i40e_config_netdev(struct i40e_vsi *vsi)
8924 {
8925         u8 brdcast[ETH_ALEN] = {0xff, 0xff, 0xff, 0xff, 0xff, 0xff};
8926         struct i40e_pf *pf = vsi->back;
8927         struct i40e_hw *hw = &pf->hw;
8928         struct i40e_netdev_priv *np;
8929         struct net_device *netdev;
8930         u8 mac_addr[ETH_ALEN];
8931         int etherdev_size;
8932
8933         etherdev_size = sizeof(struct i40e_netdev_priv);
8934         netdev = alloc_etherdev_mq(etherdev_size, vsi->alloc_queue_pairs);
8935         if (!netdev)
8936                 return -ENOMEM;
8937
8938         vsi->netdev = netdev;
8939         np = netdev_priv(netdev);
8940         np->vsi = vsi;
8941
8942         netdev->hw_enc_features |= NETIF_F_IP_CSUM       |
8943                                   NETIF_F_RXCSUM         |
8944                                   NETIF_F_GSO_UDP_TUNNEL |
8945                                   NETIF_F_GSO_GRE        |
8946                                   NETIF_F_TSO;
8947
8948         netdev->features = NETIF_F_SG                  |
8949                            NETIF_F_IP_CSUM             |
8950                            NETIF_F_SCTP_CRC            |
8951                            NETIF_F_HIGHDMA             |
8952                            NETIF_F_GSO_UDP_TUNNEL      |
8953                            NETIF_F_GSO_GRE             |
8954                            NETIF_F_HW_VLAN_CTAG_TX     |
8955                            NETIF_F_HW_VLAN_CTAG_RX     |
8956                            NETIF_F_HW_VLAN_CTAG_FILTER |
8957                            NETIF_F_IPV6_CSUM           |
8958                            NETIF_F_TSO                 |
8959                            NETIF_F_TSO_ECN             |
8960                            NETIF_F_TSO6                |
8961                            NETIF_F_RXCSUM              |
8962                            NETIF_F_RXHASH              |
8963                            0;
8964
8965         if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
8966                 netdev->features |= NETIF_F_NTUPLE;
8967
8968         /* copy netdev features into list of user selectable features */
8969         netdev->hw_features |= netdev->features;
8970
8971         if (vsi->type == I40E_VSI_MAIN) {
8972                 SET_NETDEV_DEV(netdev, &pf->pdev->dev);
8973                 ether_addr_copy(mac_addr, hw->mac.perm_addr);
8974                 /* The following steps are necessary to prevent reception
8975                  * of tagged packets - some older NVM configurations load a
8976                  * default a MAC-VLAN filter that accepts any tagged packet
8977                  * which must be replaced by a normal filter.
8978                  */
8979                 if (!i40e_rm_default_mac_filter(vsi, mac_addr)) {
8980                         spin_lock_bh(&vsi->mac_filter_list_lock);
8981                         i40e_add_filter(vsi, mac_addr,
8982                                         I40E_VLAN_ANY, false, true);
8983                         spin_unlock_bh(&vsi->mac_filter_list_lock);
8984                 }
8985         } else {
8986                 /* relate the VSI_VMDQ name to the VSI_MAIN name */
8987                 snprintf(netdev->name, IFNAMSIZ, "%sv%%d",
8988                          pf->vsi[pf->lan_vsi]->netdev->name);
8989                 random_ether_addr(mac_addr);
8990
8991                 spin_lock_bh(&vsi->mac_filter_list_lock);
8992                 i40e_add_filter(vsi, mac_addr, I40E_VLAN_ANY, false, false);
8993                 spin_unlock_bh(&vsi->mac_filter_list_lock);
8994         }
8995
8996         spin_lock_bh(&vsi->mac_filter_list_lock);
8997         i40e_add_filter(vsi, brdcast, I40E_VLAN_ANY, false, false);
8998         spin_unlock_bh(&vsi->mac_filter_list_lock);
8999
9000         ether_addr_copy(netdev->dev_addr, mac_addr);
9001         ether_addr_copy(netdev->perm_addr, mac_addr);
9002         /* vlan gets same features (except vlan offload)
9003          * after any tweaks for specific VSI types
9004          */
9005         netdev->vlan_features = netdev->features & ~(NETIF_F_HW_VLAN_CTAG_TX |
9006                                                      NETIF_F_HW_VLAN_CTAG_RX |
9007                                                    NETIF_F_HW_VLAN_CTAG_FILTER);
9008         netdev->priv_flags |= IFF_UNICAST_FLT;
9009         netdev->priv_flags |= IFF_SUPP_NOFCS;
9010         /* Setup netdev TC information */
9011         i40e_vsi_config_netdev_tc(vsi, vsi->tc_config.enabled_tc);
9012
9013         netdev->netdev_ops = &i40e_netdev_ops;
9014         netdev->watchdog_timeo = 5 * HZ;
9015         i40e_set_ethtool_ops(netdev);
9016 #ifdef I40E_FCOE
9017         i40e_fcoe_config_netdev(netdev, vsi);
9018 #endif
9019
9020         return 0;
9021 }
9022
9023 /**
9024  * i40e_vsi_delete - Delete a VSI from the switch
9025  * @vsi: the VSI being removed
9026  *
9027  * Returns 0 on success, negative value on failure
9028  **/
9029 static void i40e_vsi_delete(struct i40e_vsi *vsi)
9030 {
9031         /* remove default VSI is not allowed */
9032         if (vsi == vsi->back->vsi[vsi->back->lan_vsi])
9033                 return;
9034
9035         i40e_aq_delete_element(&vsi->back->hw, vsi->seid, NULL);
9036 }
9037
9038 /**
9039  * i40e_is_vsi_uplink_mode_veb - Check if the VSI's uplink bridge mode is VEB
9040  * @vsi: the VSI being queried
9041  *
9042  * Returns 1 if HW bridge mode is VEB and return 0 in case of VEPA mode
9043  **/
9044 int i40e_is_vsi_uplink_mode_veb(struct i40e_vsi *vsi)
9045 {
9046         struct i40e_veb *veb;
9047         struct i40e_pf *pf = vsi->back;
9048
9049         /* Uplink is not a bridge so default to VEB */
9050         if (vsi->veb_idx == I40E_NO_VEB)
9051                 return 1;
9052
9053         veb = pf->veb[vsi->veb_idx];
9054         if (!veb) {
9055                 dev_info(&pf->pdev->dev,
9056                          "There is no veb associated with the bridge\n");
9057                 return -ENOENT;
9058         }
9059
9060         /* Uplink is a bridge in VEPA mode */
9061         if (veb->bridge_mode & BRIDGE_MODE_VEPA) {
9062                 return 0;
9063         } else {
9064                 /* Uplink is a bridge in VEB mode */
9065                 return 1;
9066         }
9067
9068         /* VEPA is now default bridge, so return 0 */
9069         return 0;
9070 }
9071
9072 /**
9073  * i40e_add_vsi - Add a VSI to the switch
9074  * @vsi: the VSI being configured
9075  *
9076  * This initializes a VSI context depending on the VSI type to be added and
9077  * passes it down to the add_vsi aq command.
9078  **/
9079 static int i40e_add_vsi(struct i40e_vsi *vsi)
9080 {
9081         int ret = -ENODEV;
9082         u8 laa_macaddr[ETH_ALEN];
9083         bool found_laa_mac_filter = false;
9084         struct i40e_pf *pf = vsi->back;
9085         struct i40e_hw *hw = &pf->hw;
9086         struct i40e_vsi_context ctxt;
9087         struct i40e_mac_filter *f, *ftmp;
9088
9089         u8 enabled_tc = 0x1; /* TC0 enabled */
9090         int f_count = 0;
9091
9092         memset(&ctxt, 0, sizeof(ctxt));
9093         switch (vsi->type) {
9094         case I40E_VSI_MAIN:
9095                 /* The PF's main VSI is already setup as part of the
9096                  * device initialization, so we'll not bother with
9097                  * the add_vsi call, but we will retrieve the current
9098                  * VSI context.
9099                  */
9100                 ctxt.seid = pf->main_vsi_seid;
9101                 ctxt.pf_num = pf->hw.pf_id;
9102                 ctxt.vf_num = 0;
9103                 ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
9104                 ctxt.flags = I40E_AQ_VSI_TYPE_PF;
9105                 if (ret) {
9106                         dev_info(&pf->pdev->dev,
9107                                  "couldn't get PF vsi config, err %s aq_err %s\n",
9108                                  i40e_stat_str(&pf->hw, ret),
9109                                  i40e_aq_str(&pf->hw,
9110                                              pf->hw.aq.asq_last_status));
9111                         return -ENOENT;
9112                 }
9113                 vsi->info = ctxt.info;
9114                 vsi->info.valid_sections = 0;
9115
9116                 vsi->seid = ctxt.seid;
9117                 vsi->id = ctxt.vsi_number;
9118
9119                 enabled_tc = i40e_pf_get_tc_map(pf);
9120
9121                 /* MFP mode setup queue map and update VSI */
9122                 if ((pf->flags & I40E_FLAG_MFP_ENABLED) &&
9123                     !(pf->hw.func_caps.iscsi)) { /* NIC type PF */
9124                         memset(&ctxt, 0, sizeof(ctxt));
9125                         ctxt.seid = pf->main_vsi_seid;
9126                         ctxt.pf_num = pf->hw.pf_id;
9127                         ctxt.vf_num = 0;
9128                         i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, false);
9129                         ret = i40e_aq_update_vsi_params(hw, &ctxt, NULL);
9130                         if (ret) {
9131                                 dev_info(&pf->pdev->dev,
9132                                          "update vsi failed, err %s aq_err %s\n",
9133                                          i40e_stat_str(&pf->hw, ret),
9134                                          i40e_aq_str(&pf->hw,
9135                                                     pf->hw.aq.asq_last_status));
9136                                 ret = -ENOENT;
9137                                 goto err;
9138                         }
9139                         /* update the local VSI info queue map */
9140                         i40e_vsi_update_queue_map(vsi, &ctxt);
9141                         vsi->info.valid_sections = 0;
9142                 } else {
9143                         /* Default/Main VSI is only enabled for TC0
9144                          * reconfigure it to enable all TCs that are
9145                          * available on the port in SFP mode.
9146                          * For MFP case the iSCSI PF would use this
9147                          * flow to enable LAN+iSCSI TC.
9148                          */
9149                         ret = i40e_vsi_config_tc(vsi, enabled_tc);
9150                         if (ret) {
9151                                 dev_info(&pf->pdev->dev,
9152                                          "failed to configure TCs for main VSI tc_map 0x%08x, err %s aq_err %s\n",
9153                                          enabled_tc,
9154                                          i40e_stat_str(&pf->hw, ret),
9155                                          i40e_aq_str(&pf->hw,
9156                                                     pf->hw.aq.asq_last_status));
9157                                 ret = -ENOENT;
9158                         }
9159                 }
9160                 break;
9161
9162         case I40E_VSI_FDIR:
9163                 ctxt.pf_num = hw->pf_id;
9164                 ctxt.vf_num = 0;
9165                 ctxt.uplink_seid = vsi->uplink_seid;
9166                 ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
9167                 ctxt.flags = I40E_AQ_VSI_TYPE_PF;
9168                 if ((pf->flags & I40E_FLAG_VEB_MODE_ENABLED) &&
9169                     (i40e_is_vsi_uplink_mode_veb(vsi))) {
9170                         ctxt.info.valid_sections |=
9171                              cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
9172                         ctxt.info.switch_id =
9173                            cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
9174                 }
9175                 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
9176                 break;
9177
9178         case I40E_VSI_VMDQ2:
9179                 ctxt.pf_num = hw->pf_id;
9180                 ctxt.vf_num = 0;
9181                 ctxt.uplink_seid = vsi->uplink_seid;
9182                 ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
9183                 ctxt.flags = I40E_AQ_VSI_TYPE_VMDQ2;
9184
9185                 /* This VSI is connected to VEB so the switch_id
9186                  * should be set to zero by default.
9187                  */
9188                 if (i40e_is_vsi_uplink_mode_veb(vsi)) {
9189                         ctxt.info.valid_sections |=
9190                                 cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
9191                         ctxt.info.switch_id =
9192                                 cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
9193                 }
9194
9195                 /* Setup the VSI tx/rx queue map for TC0 only for now */
9196                 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
9197                 break;
9198
9199         case I40E_VSI_SRIOV:
9200                 ctxt.pf_num = hw->pf_id;
9201                 ctxt.vf_num = vsi->vf_id + hw->func_caps.vf_base_id;
9202                 ctxt.uplink_seid = vsi->uplink_seid;
9203                 ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
9204                 ctxt.flags = I40E_AQ_VSI_TYPE_VF;
9205
9206                 /* This VSI is connected to VEB so the switch_id
9207                  * should be set to zero by default.
9208                  */
9209                 if (i40e_is_vsi_uplink_mode_veb(vsi)) {
9210                         ctxt.info.valid_sections |=
9211                                 cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
9212                         ctxt.info.switch_id =
9213                                 cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
9214                 }
9215
9216                 ctxt.info.valid_sections |= cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
9217                 ctxt.info.port_vlan_flags |= I40E_AQ_VSI_PVLAN_MODE_ALL;
9218                 if (pf->vf[vsi->vf_id].spoofchk) {
9219                         ctxt.info.valid_sections |=
9220                                 cpu_to_le16(I40E_AQ_VSI_PROP_SECURITY_VALID);
9221                         ctxt.info.sec_flags |=
9222                                 (I40E_AQ_VSI_SEC_FLAG_ENABLE_VLAN_CHK |
9223                                  I40E_AQ_VSI_SEC_FLAG_ENABLE_MAC_CHK);
9224                 }
9225                 /* Setup the VSI tx/rx queue map for TC0 only for now */
9226                 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
9227                 break;
9228
9229 #ifdef I40E_FCOE
9230         case I40E_VSI_FCOE:
9231                 ret = i40e_fcoe_vsi_init(vsi, &ctxt);
9232                 if (ret) {
9233                         dev_info(&pf->pdev->dev, "failed to initialize FCoE VSI\n");
9234                         return ret;
9235                 }
9236                 break;
9237
9238 #endif /* I40E_FCOE */
9239         default:
9240                 return -ENODEV;
9241         }
9242
9243         if (vsi->type != I40E_VSI_MAIN) {
9244                 ret = i40e_aq_add_vsi(hw, &ctxt, NULL);
9245                 if (ret) {
9246                         dev_info(&vsi->back->pdev->dev,
9247                                  "add vsi failed, err %s aq_err %s\n",
9248                                  i40e_stat_str(&pf->hw, ret),
9249                                  i40e_aq_str(&pf->hw,
9250                                              pf->hw.aq.asq_last_status));
9251                         ret = -ENOENT;
9252                         goto err;
9253                 }
9254                 vsi->info = ctxt.info;
9255                 vsi->info.valid_sections = 0;
9256                 vsi->seid = ctxt.seid;
9257                 vsi->id = ctxt.vsi_number;
9258         }
9259
9260         spin_lock_bh(&vsi->mac_filter_list_lock);
9261         /* If macvlan filters already exist, force them to get loaded */
9262         list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
9263                 f->changed = true;
9264                 f_count++;
9265
9266                 /* Expected to have only one MAC filter entry for LAA in list */
9267                 if (f->is_laa && vsi->type == I40E_VSI_MAIN) {
9268                         ether_addr_copy(laa_macaddr, f->macaddr);
9269                         found_laa_mac_filter = true;
9270                 }
9271         }
9272         spin_unlock_bh(&vsi->mac_filter_list_lock);
9273
9274         if (found_laa_mac_filter) {
9275                 struct i40e_aqc_remove_macvlan_element_data element;
9276
9277                 memset(&element, 0, sizeof(element));
9278                 ether_addr_copy(element.mac_addr, laa_macaddr);
9279                 element.flags = I40E_AQC_MACVLAN_DEL_PERFECT_MATCH;
9280                 ret = i40e_aq_remove_macvlan(hw, vsi->seid,
9281                                              &element, 1, NULL);
9282                 if (ret) {
9283                         /* some older FW has a different default */
9284                         element.flags |=
9285                                        I40E_AQC_MACVLAN_DEL_IGNORE_VLAN;
9286                         i40e_aq_remove_macvlan(hw, vsi->seid,
9287                                                &element, 1, NULL);
9288                 }
9289
9290                 i40e_aq_mac_address_write(hw,
9291                                           I40E_AQC_WRITE_TYPE_LAA_WOL,
9292                                           laa_macaddr, NULL);
9293         }
9294
9295         if (f_count) {
9296                 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
9297                 pf->flags |= I40E_FLAG_FILTER_SYNC;
9298         }
9299
9300         /* Update VSI BW information */
9301         ret = i40e_vsi_get_bw_info(vsi);
9302         if (ret) {
9303                 dev_info(&pf->pdev->dev,
9304                          "couldn't get vsi bw info, err %s aq_err %s\n",
9305                          i40e_stat_str(&pf->hw, ret),
9306                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
9307                 /* VSI is already added so not tearing that up */
9308                 ret = 0;
9309         }
9310
9311 err:
9312         return ret;
9313 }
9314
9315 /**
9316  * i40e_vsi_release - Delete a VSI and free its resources
9317  * @vsi: the VSI being removed
9318  *
9319  * Returns 0 on success or < 0 on error
9320  **/
9321 int i40e_vsi_release(struct i40e_vsi *vsi)
9322 {
9323         struct i40e_mac_filter *f, *ftmp;
9324         struct i40e_veb *veb = NULL;
9325         struct i40e_pf *pf;
9326         u16 uplink_seid;
9327         int i, n;
9328
9329         pf = vsi->back;
9330
9331         /* release of a VEB-owner or last VSI is not allowed */
9332         if (vsi->flags & I40E_VSI_FLAG_VEB_OWNER) {
9333                 dev_info(&pf->pdev->dev, "VSI %d has existing VEB %d\n",
9334                          vsi->seid, vsi->uplink_seid);
9335                 return -ENODEV;
9336         }
9337         if (vsi == pf->vsi[pf->lan_vsi] &&
9338             !test_bit(__I40E_DOWN, &pf->state)) {
9339                 dev_info(&pf->pdev->dev, "Can't remove PF VSI\n");
9340                 return -ENODEV;
9341         }
9342
9343         uplink_seid = vsi->uplink_seid;
9344         if (vsi->type != I40E_VSI_SRIOV) {
9345                 if (vsi->netdev_registered) {
9346                         vsi->netdev_registered = false;
9347                         if (vsi->netdev) {
9348                                 /* results in a call to i40e_close() */
9349                                 unregister_netdev(vsi->netdev);
9350                         }
9351                 } else {
9352                         i40e_vsi_close(vsi);
9353                 }
9354                 i40e_vsi_disable_irq(vsi);
9355         }
9356
9357         spin_lock_bh(&vsi->mac_filter_list_lock);
9358         list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list)
9359                 i40e_del_filter(vsi, f->macaddr, f->vlan,
9360                                 f->is_vf, f->is_netdev);
9361         spin_unlock_bh(&vsi->mac_filter_list_lock);
9362
9363         i40e_sync_vsi_filters(vsi);
9364
9365         i40e_vsi_delete(vsi);
9366         i40e_vsi_free_q_vectors(vsi);
9367         if (vsi->netdev) {
9368                 free_netdev(vsi->netdev);
9369                 vsi->netdev = NULL;
9370         }
9371         i40e_vsi_clear_rings(vsi);
9372         i40e_vsi_clear(vsi);
9373
9374         /* If this was the last thing on the VEB, except for the
9375          * controlling VSI, remove the VEB, which puts the controlling
9376          * VSI onto the next level down in the switch.
9377          *
9378          * Well, okay, there's one more exception here: don't remove
9379          * the orphan VEBs yet.  We'll wait for an explicit remove request
9380          * from up the network stack.
9381          */
9382         for (n = 0, i = 0; i < pf->num_alloc_vsi; i++) {
9383                 if (pf->vsi[i] &&
9384                     pf->vsi[i]->uplink_seid == uplink_seid &&
9385                     (pf->vsi[i]->flags & I40E_VSI_FLAG_VEB_OWNER) == 0) {
9386                         n++;      /* count the VSIs */
9387                 }
9388         }
9389         for (i = 0; i < I40E_MAX_VEB; i++) {
9390                 if (!pf->veb[i])
9391                         continue;
9392                 if (pf->veb[i]->uplink_seid == uplink_seid)
9393                         n++;     /* count the VEBs */
9394                 if (pf->veb[i]->seid == uplink_seid)
9395                         veb = pf->veb[i];
9396         }
9397         if (n == 0 && veb && veb->uplink_seid != 0)
9398                 i40e_veb_release(veb);
9399
9400         return 0;
9401 }
9402
9403 /**
9404  * i40e_vsi_setup_vectors - Set up the q_vectors for the given VSI
9405  * @vsi: ptr to the VSI
9406  *
9407  * This should only be called after i40e_vsi_mem_alloc() which allocates the
9408  * corresponding SW VSI structure and initializes num_queue_pairs for the
9409  * newly allocated VSI.
9410  *
9411  * Returns 0 on success or negative on failure
9412  **/
9413 static int i40e_vsi_setup_vectors(struct i40e_vsi *vsi)
9414 {
9415         int ret = -ENOENT;
9416         struct i40e_pf *pf = vsi->back;
9417
9418         if (vsi->q_vectors[0]) {
9419                 dev_info(&pf->pdev->dev, "VSI %d has existing q_vectors\n",
9420                          vsi->seid);
9421                 return -EEXIST;
9422         }
9423
9424         if (vsi->base_vector) {
9425                 dev_info(&pf->pdev->dev, "VSI %d has non-zero base vector %d\n",
9426                          vsi->seid, vsi->base_vector);
9427                 return -EEXIST;
9428         }
9429
9430         ret = i40e_vsi_alloc_q_vectors(vsi);
9431         if (ret) {
9432                 dev_info(&pf->pdev->dev,
9433                          "failed to allocate %d q_vector for VSI %d, ret=%d\n",
9434                          vsi->num_q_vectors, vsi->seid, ret);
9435                 vsi->num_q_vectors = 0;
9436                 goto vector_setup_out;
9437         }
9438
9439         /* In Legacy mode, we do not have to get any other vector since we
9440          * piggyback on the misc/ICR0 for queue interrupts.
9441         */
9442         if (!(pf->flags & I40E_FLAG_MSIX_ENABLED))
9443                 return ret;
9444         if (vsi->num_q_vectors)
9445                 vsi->base_vector = i40e_get_lump(pf, pf->irq_pile,
9446                                                  vsi->num_q_vectors, vsi->idx);
9447         if (vsi->base_vector < 0) {
9448                 dev_info(&pf->pdev->dev,
9449                          "failed to get tracking for %d vectors for VSI %d, err=%d\n",
9450                          vsi->num_q_vectors, vsi->seid, vsi->base_vector);
9451                 i40e_vsi_free_q_vectors(vsi);
9452                 ret = -ENOENT;
9453                 goto vector_setup_out;
9454         }
9455
9456 vector_setup_out:
9457         return ret;
9458 }
9459
9460 /**
9461  * i40e_vsi_reinit_setup - return and reallocate resources for a VSI
9462  * @vsi: pointer to the vsi.
9463  *
9464  * This re-allocates a vsi's queue resources.
9465  *
9466  * Returns pointer to the successfully allocated and configured VSI sw struct
9467  * on success, otherwise returns NULL on failure.
9468  **/
9469 static struct i40e_vsi *i40e_vsi_reinit_setup(struct i40e_vsi *vsi)
9470 {
9471         struct i40e_pf *pf = vsi->back;
9472         u8 enabled_tc;
9473         int ret;
9474
9475         i40e_put_lump(pf->qp_pile, vsi->base_queue, vsi->idx);
9476         i40e_vsi_clear_rings(vsi);
9477
9478         i40e_vsi_free_arrays(vsi, false);
9479         i40e_set_num_rings_in_vsi(vsi);
9480         ret = i40e_vsi_alloc_arrays(vsi, false);
9481         if (ret)
9482                 goto err_vsi;
9483
9484         ret = i40e_get_lump(pf, pf->qp_pile, vsi->alloc_queue_pairs, vsi->idx);
9485         if (ret < 0) {
9486                 dev_info(&pf->pdev->dev,
9487                          "failed to get tracking for %d queues for VSI %d err %d\n",
9488                          vsi->alloc_queue_pairs, vsi->seid, ret);
9489                 goto err_vsi;
9490         }
9491         vsi->base_queue = ret;
9492
9493         /* Update the FW view of the VSI. Force a reset of TC and queue
9494          * layout configurations.
9495          */
9496         enabled_tc = pf->vsi[pf->lan_vsi]->tc_config.enabled_tc;
9497         pf->vsi[pf->lan_vsi]->tc_config.enabled_tc = 0;
9498         pf->vsi[pf->lan_vsi]->seid = pf->main_vsi_seid;
9499         i40e_vsi_config_tc(pf->vsi[pf->lan_vsi], enabled_tc);
9500
9501         /* assign it some queues */
9502         ret = i40e_alloc_rings(vsi);
9503         if (ret)
9504                 goto err_rings;
9505
9506         /* map all of the rings to the q_vectors */
9507         i40e_vsi_map_rings_to_vectors(vsi);
9508         return vsi;
9509
9510 err_rings:
9511         i40e_vsi_free_q_vectors(vsi);
9512         if (vsi->netdev_registered) {
9513                 vsi->netdev_registered = false;
9514                 unregister_netdev(vsi->netdev);
9515                 free_netdev(vsi->netdev);
9516                 vsi->netdev = NULL;
9517         }
9518         i40e_aq_delete_element(&pf->hw, vsi->seid, NULL);
9519 err_vsi:
9520         i40e_vsi_clear(vsi);
9521         return NULL;
9522 }
9523
9524 /**
9525  * i40e_vsi_setup - Set up a VSI by a given type
9526  * @pf: board private structure
9527  * @type: VSI type
9528  * @uplink_seid: the switch element to link to
9529  * @param1: usage depends upon VSI type. For VF types, indicates VF id
9530  *
9531  * This allocates the sw VSI structure and its queue resources, then add a VSI
9532  * to the identified VEB.
9533  *
9534  * Returns pointer to the successfully allocated and configure VSI sw struct on
9535  * success, otherwise returns NULL on failure.
9536  **/
9537 struct i40e_vsi *i40e_vsi_setup(struct i40e_pf *pf, u8 type,
9538                                 u16 uplink_seid, u32 param1)
9539 {
9540         struct i40e_vsi *vsi = NULL;
9541         struct i40e_veb *veb = NULL;
9542         int ret, i;
9543         int v_idx;
9544
9545         /* The requested uplink_seid must be either
9546          *     - the PF's port seid
9547          *              no VEB is needed because this is the PF
9548          *              or this is a Flow Director special case VSI
9549          *     - seid of an existing VEB
9550          *     - seid of a VSI that owns an existing VEB
9551          *     - seid of a VSI that doesn't own a VEB
9552          *              a new VEB is created and the VSI becomes the owner
9553          *     - seid of the PF VSI, which is what creates the first VEB
9554          *              this is a special case of the previous
9555          *
9556          * Find which uplink_seid we were given and create a new VEB if needed
9557          */
9558         for (i = 0; i < I40E_MAX_VEB; i++) {
9559                 if (pf->veb[i] && pf->veb[i]->seid == uplink_seid) {
9560                         veb = pf->veb[i];
9561                         break;
9562                 }
9563         }
9564
9565         if (!veb && uplink_seid != pf->mac_seid) {
9566
9567                 for (i = 0; i < pf->num_alloc_vsi; i++) {
9568                         if (pf->vsi[i] && pf->vsi[i]->seid == uplink_seid) {
9569                                 vsi = pf->vsi[i];
9570                                 break;
9571                         }
9572                 }
9573                 if (!vsi) {
9574                         dev_info(&pf->pdev->dev, "no such uplink_seid %d\n",
9575                                  uplink_seid);
9576                         return NULL;
9577                 }
9578
9579                 if (vsi->uplink_seid == pf->mac_seid)
9580                         veb = i40e_veb_setup(pf, 0, pf->mac_seid, vsi->seid,
9581                                              vsi->tc_config.enabled_tc);
9582                 else if ((vsi->flags & I40E_VSI_FLAG_VEB_OWNER) == 0)
9583                         veb = i40e_veb_setup(pf, 0, vsi->uplink_seid, vsi->seid,
9584                                              vsi->tc_config.enabled_tc);
9585                 if (veb) {
9586                         if (vsi->seid != pf->vsi[pf->lan_vsi]->seid) {
9587                                 dev_info(&vsi->back->pdev->dev,
9588                                          "New VSI creation error, uplink seid of LAN VSI expected.\n");
9589                                 return NULL;
9590                         }
9591                         /* We come up by default in VEPA mode if SRIOV is not
9592                          * already enabled, in which case we can't force VEPA
9593                          * mode.
9594                          */
9595                         if (!(pf->flags & I40E_FLAG_VEB_MODE_ENABLED)) {
9596                                 veb->bridge_mode = BRIDGE_MODE_VEPA;
9597                                 pf->flags &= ~I40E_FLAG_VEB_MODE_ENABLED;
9598                         }
9599                         i40e_config_bridge_mode(veb);
9600                 }
9601                 for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
9602                         if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
9603                                 veb = pf->veb[i];
9604                 }
9605                 if (!veb) {
9606                         dev_info(&pf->pdev->dev, "couldn't add VEB\n");
9607                         return NULL;
9608                 }
9609
9610                 vsi->flags |= I40E_VSI_FLAG_VEB_OWNER;
9611                 uplink_seid = veb->seid;
9612         }
9613
9614         /* get vsi sw struct */
9615         v_idx = i40e_vsi_mem_alloc(pf, type);
9616         if (v_idx < 0)
9617                 goto err_alloc;
9618         vsi = pf->vsi[v_idx];
9619         if (!vsi)
9620                 goto err_alloc;
9621         vsi->type = type;
9622         vsi->veb_idx = (veb ? veb->idx : I40E_NO_VEB);
9623
9624         if (type == I40E_VSI_MAIN)
9625                 pf->lan_vsi = v_idx;
9626         else if (type == I40E_VSI_SRIOV)
9627                 vsi->vf_id = param1;
9628         /* assign it some queues */
9629         ret = i40e_get_lump(pf, pf->qp_pile, vsi->alloc_queue_pairs,
9630                                 vsi->idx);
9631         if (ret < 0) {
9632                 dev_info(&pf->pdev->dev,
9633                          "failed to get tracking for %d queues for VSI %d err=%d\n",
9634                          vsi->alloc_queue_pairs, vsi->seid, ret);
9635                 goto err_vsi;
9636         }
9637         vsi->base_queue = ret;
9638
9639         /* get a VSI from the hardware */
9640         vsi->uplink_seid = uplink_seid;
9641         ret = i40e_add_vsi(vsi);
9642         if (ret)
9643                 goto err_vsi;
9644
9645         switch (vsi->type) {
9646         /* setup the netdev if needed */
9647         case I40E_VSI_MAIN:
9648         case I40E_VSI_VMDQ2:
9649         case I40E_VSI_FCOE:
9650                 ret = i40e_config_netdev(vsi);
9651                 if (ret)
9652                         goto err_netdev;
9653                 ret = register_netdev(vsi->netdev);
9654                 if (ret)
9655                         goto err_netdev;
9656                 vsi->netdev_registered = true;
9657                 netif_carrier_off(vsi->netdev);
9658 #ifdef CONFIG_I40E_DCB
9659                 /* Setup DCB netlink interface */
9660                 i40e_dcbnl_setup(vsi);
9661 #endif /* CONFIG_I40E_DCB */
9662                 /* fall through */
9663
9664         case I40E_VSI_FDIR:
9665                 /* set up vectors and rings if needed */
9666                 ret = i40e_vsi_setup_vectors(vsi);
9667                 if (ret)
9668                         goto err_msix;
9669
9670                 ret = i40e_alloc_rings(vsi);
9671                 if (ret)
9672                         goto err_rings;
9673
9674                 /* map all of the rings to the q_vectors */
9675                 i40e_vsi_map_rings_to_vectors(vsi);
9676
9677                 i40e_vsi_reset_stats(vsi);
9678                 break;
9679
9680         default:
9681                 /* no netdev or rings for the other VSI types */
9682                 break;
9683         }
9684
9685         if ((pf->flags & I40E_FLAG_RSS_AQ_CAPABLE) &&
9686             (vsi->type == I40E_VSI_VMDQ2)) {
9687                 ret = i40e_vsi_config_rss(vsi);
9688         }
9689         return vsi;
9690
9691 err_rings:
9692         i40e_vsi_free_q_vectors(vsi);
9693 err_msix:
9694         if (vsi->netdev_registered) {
9695                 vsi->netdev_registered = false;
9696                 unregister_netdev(vsi->netdev);
9697                 free_netdev(vsi->netdev);
9698                 vsi->netdev = NULL;
9699         }
9700 err_netdev:
9701         i40e_aq_delete_element(&pf->hw, vsi->seid, NULL);
9702 err_vsi:
9703         i40e_vsi_clear(vsi);
9704 err_alloc:
9705         return NULL;
9706 }
9707
9708 /**
9709  * i40e_veb_get_bw_info - Query VEB BW information
9710  * @veb: the veb to query
9711  *
9712  * Query the Tx scheduler BW configuration data for given VEB
9713  **/
9714 static int i40e_veb_get_bw_info(struct i40e_veb *veb)
9715 {
9716         struct i40e_aqc_query_switching_comp_ets_config_resp ets_data;
9717         struct i40e_aqc_query_switching_comp_bw_config_resp bw_data;
9718         struct i40e_pf *pf = veb->pf;
9719         struct i40e_hw *hw = &pf->hw;
9720         u32 tc_bw_max;
9721         int ret = 0;
9722         int i;
9723
9724         ret = i40e_aq_query_switch_comp_bw_config(hw, veb->seid,
9725                                                   &bw_data, NULL);
9726         if (ret) {
9727                 dev_info(&pf->pdev->dev,
9728                          "query veb bw config failed, err %s aq_err %s\n",
9729                          i40e_stat_str(&pf->hw, ret),
9730                          i40e_aq_str(&pf->hw, hw->aq.asq_last_status));
9731                 goto out;
9732         }
9733
9734         ret = i40e_aq_query_switch_comp_ets_config(hw, veb->seid,
9735                                                    &ets_data, NULL);
9736         if (ret) {
9737                 dev_info(&pf->pdev->dev,
9738                          "query veb bw ets config failed, err %s aq_err %s\n",
9739                          i40e_stat_str(&pf->hw, ret),
9740                          i40e_aq_str(&pf->hw, hw->aq.asq_last_status));
9741                 goto out;
9742         }
9743
9744         veb->bw_limit = le16_to_cpu(ets_data.port_bw_limit);
9745         veb->bw_max_quanta = ets_data.tc_bw_max;
9746         veb->is_abs_credits = bw_data.absolute_credits_enable;
9747         veb->enabled_tc = ets_data.tc_valid_bits;
9748         tc_bw_max = le16_to_cpu(bw_data.tc_bw_max[0]) |
9749                     (le16_to_cpu(bw_data.tc_bw_max[1]) << 16);
9750         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
9751                 veb->bw_tc_share_credits[i] = bw_data.tc_bw_share_credits[i];
9752                 veb->bw_tc_limit_credits[i] =
9753                                         le16_to_cpu(bw_data.tc_bw_limits[i]);
9754                 veb->bw_tc_max_quanta[i] = ((tc_bw_max >> (i*4)) & 0x7);
9755         }
9756
9757 out:
9758         return ret;
9759 }
9760
9761 /**
9762  * i40e_veb_mem_alloc - Allocates the next available struct veb in the PF
9763  * @pf: board private structure
9764  *
9765  * On error: returns error code (negative)
9766  * On success: returns vsi index in PF (positive)
9767  **/
9768 static int i40e_veb_mem_alloc(struct i40e_pf *pf)
9769 {
9770         int ret = -ENOENT;
9771         struct i40e_veb *veb;
9772         int i;
9773
9774         /* Need to protect the allocation of switch elements at the PF level */
9775         mutex_lock(&pf->switch_mutex);
9776
9777         /* VEB list may be fragmented if VEB creation/destruction has
9778          * been happening.  We can afford to do a quick scan to look
9779          * for any free slots in the list.
9780          *
9781          * find next empty veb slot, looping back around if necessary
9782          */
9783         i = 0;
9784         while ((i < I40E_MAX_VEB) && (pf->veb[i] != NULL))
9785                 i++;
9786         if (i >= I40E_MAX_VEB) {
9787                 ret = -ENOMEM;
9788                 goto err_alloc_veb;  /* out of VEB slots! */
9789         }
9790
9791         veb = kzalloc(sizeof(*veb), GFP_KERNEL);
9792         if (!veb) {
9793                 ret = -ENOMEM;
9794                 goto err_alloc_veb;
9795         }
9796         veb->pf = pf;
9797         veb->idx = i;
9798         veb->enabled_tc = 1;
9799
9800         pf->veb[i] = veb;
9801         ret = i;
9802 err_alloc_veb:
9803         mutex_unlock(&pf->switch_mutex);
9804         return ret;
9805 }
9806
9807 /**
9808  * i40e_switch_branch_release - Delete a branch of the switch tree
9809  * @branch: where to start deleting
9810  *
9811  * This uses recursion to find the tips of the branch to be
9812  * removed, deleting until we get back to and can delete this VEB.
9813  **/
9814 static void i40e_switch_branch_release(struct i40e_veb *branch)
9815 {
9816         struct i40e_pf *pf = branch->pf;
9817         u16 branch_seid = branch->seid;
9818         u16 veb_idx = branch->idx;
9819         int i;
9820
9821         /* release any VEBs on this VEB - RECURSION */
9822         for (i = 0; i < I40E_MAX_VEB; i++) {
9823                 if (!pf->veb[i])
9824                         continue;
9825                 if (pf->veb[i]->uplink_seid == branch->seid)
9826                         i40e_switch_branch_release(pf->veb[i]);
9827         }
9828
9829         /* Release the VSIs on this VEB, but not the owner VSI.
9830          *
9831          * NOTE: Removing the last VSI on a VEB has the SIDE EFFECT of removing
9832          *       the VEB itself, so don't use (*branch) after this loop.
9833          */
9834         for (i = 0; i < pf->num_alloc_vsi; i++) {
9835                 if (!pf->vsi[i])
9836                         continue;
9837                 if (pf->vsi[i]->uplink_seid == branch_seid &&
9838                    (pf->vsi[i]->flags & I40E_VSI_FLAG_VEB_OWNER) == 0) {
9839                         i40e_vsi_release(pf->vsi[i]);
9840                 }
9841         }
9842
9843         /* There's one corner case where the VEB might not have been
9844          * removed, so double check it here and remove it if needed.
9845          * This case happens if the veb was created from the debugfs
9846          * commands and no VSIs were added to it.
9847          */
9848         if (pf->veb[veb_idx])
9849                 i40e_veb_release(pf->veb[veb_idx]);
9850 }
9851
9852 /**
9853  * i40e_veb_clear - remove veb struct
9854  * @veb: the veb to remove
9855  **/
9856 static void i40e_veb_clear(struct i40e_veb *veb)
9857 {
9858         if (!veb)
9859                 return;
9860
9861         if (veb->pf) {
9862                 struct i40e_pf *pf = veb->pf;
9863
9864                 mutex_lock(&pf->switch_mutex);
9865                 if (pf->veb[veb->idx] == veb)
9866                         pf->veb[veb->idx] = NULL;
9867                 mutex_unlock(&pf->switch_mutex);
9868         }
9869
9870         kfree(veb);
9871 }
9872
9873 /**
9874  * i40e_veb_release - Delete a VEB and free its resources
9875  * @veb: the VEB being removed
9876  **/
9877 void i40e_veb_release(struct i40e_veb *veb)
9878 {
9879         struct i40e_vsi *vsi = NULL;
9880         struct i40e_pf *pf;
9881         int i, n = 0;
9882
9883         pf = veb->pf;
9884
9885         /* find the remaining VSI and check for extras */
9886         for (i = 0; i < pf->num_alloc_vsi; i++) {
9887                 if (pf->vsi[i] && pf->vsi[i]->uplink_seid == veb->seid) {
9888                         n++;
9889                         vsi = pf->vsi[i];
9890                 }
9891         }
9892         if (n != 1) {
9893                 dev_info(&pf->pdev->dev,
9894                          "can't remove VEB %d with %d VSIs left\n",
9895                          veb->seid, n);
9896                 return;
9897         }
9898
9899         /* move the remaining VSI to uplink veb */
9900         vsi->flags &= ~I40E_VSI_FLAG_VEB_OWNER;
9901         if (veb->uplink_seid) {
9902                 vsi->uplink_seid = veb->uplink_seid;
9903                 if (veb->uplink_seid == pf->mac_seid)
9904                         vsi->veb_idx = I40E_NO_VEB;
9905                 else
9906                         vsi->veb_idx = veb->veb_idx;
9907         } else {
9908                 /* floating VEB */
9909                 vsi->uplink_seid = pf->vsi[pf->lan_vsi]->uplink_seid;
9910                 vsi->veb_idx = pf->vsi[pf->lan_vsi]->veb_idx;
9911         }
9912
9913         i40e_aq_delete_element(&pf->hw, veb->seid, NULL);
9914         i40e_veb_clear(veb);
9915 }
9916
9917 /**
9918  * i40e_add_veb - create the VEB in the switch
9919  * @veb: the VEB to be instantiated
9920  * @vsi: the controlling VSI
9921  **/
9922 static int i40e_add_veb(struct i40e_veb *veb, struct i40e_vsi *vsi)
9923 {
9924         struct i40e_pf *pf = veb->pf;
9925         bool is_default = veb->pf->cur_promisc;
9926         bool is_cloud = false;
9927         int ret;
9928
9929         /* get a VEB from the hardware */
9930         ret = i40e_aq_add_veb(&pf->hw, veb->uplink_seid, vsi->seid,
9931                               veb->enabled_tc, is_default,
9932                               is_cloud, &veb->seid, NULL);
9933         if (ret) {
9934                 dev_info(&pf->pdev->dev,
9935                          "couldn't add VEB, err %s aq_err %s\n",
9936                          i40e_stat_str(&pf->hw, ret),
9937                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
9938                 return -EPERM;
9939         }
9940
9941         /* get statistics counter */
9942         ret = i40e_aq_get_veb_parameters(&pf->hw, veb->seid, NULL, NULL,
9943                                          &veb->stats_idx, NULL, NULL, NULL);
9944         if (ret) {
9945                 dev_info(&pf->pdev->dev,
9946                          "couldn't get VEB statistics idx, err %s aq_err %s\n",
9947                          i40e_stat_str(&pf->hw, ret),
9948                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
9949                 return -EPERM;
9950         }
9951         ret = i40e_veb_get_bw_info(veb);
9952         if (ret) {
9953                 dev_info(&pf->pdev->dev,
9954                          "couldn't get VEB bw info, err %s aq_err %s\n",
9955                          i40e_stat_str(&pf->hw, ret),
9956                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
9957                 i40e_aq_delete_element(&pf->hw, veb->seid, NULL);
9958                 return -ENOENT;
9959         }
9960
9961         vsi->uplink_seid = veb->seid;
9962         vsi->veb_idx = veb->idx;
9963         vsi->flags |= I40E_VSI_FLAG_VEB_OWNER;
9964
9965         return 0;
9966 }
9967
9968 /**
9969  * i40e_veb_setup - Set up a VEB
9970  * @pf: board private structure
9971  * @flags: VEB setup flags
9972  * @uplink_seid: the switch element to link to
9973  * @vsi_seid: the initial VSI seid
9974  * @enabled_tc: Enabled TC bit-map
9975  *
9976  * This allocates the sw VEB structure and links it into the switch
9977  * It is possible and legal for this to be a duplicate of an already
9978  * existing VEB.  It is also possible for both uplink and vsi seids
9979  * to be zero, in order to create a floating VEB.
9980  *
9981  * Returns pointer to the successfully allocated VEB sw struct on
9982  * success, otherwise returns NULL on failure.
9983  **/
9984 struct i40e_veb *i40e_veb_setup(struct i40e_pf *pf, u16 flags,
9985                                 u16 uplink_seid, u16 vsi_seid,
9986                                 u8 enabled_tc)
9987 {
9988         struct i40e_veb *veb, *uplink_veb = NULL;
9989         int vsi_idx, veb_idx;
9990         int ret;
9991
9992         /* if one seid is 0, the other must be 0 to create a floating relay */
9993         if ((uplink_seid == 0 || vsi_seid == 0) &&
9994             (uplink_seid + vsi_seid != 0)) {
9995                 dev_info(&pf->pdev->dev,
9996                          "one, not both seid's are 0: uplink=%d vsi=%d\n",
9997                          uplink_seid, vsi_seid);
9998                 return NULL;
9999         }
10000
10001         /* make sure there is such a vsi and uplink */
10002         for (vsi_idx = 0; vsi_idx < pf->num_alloc_vsi; vsi_idx++)
10003                 if (pf->vsi[vsi_idx] && pf->vsi[vsi_idx]->seid == vsi_seid)
10004                         break;
10005         if (vsi_idx >= pf->num_alloc_vsi && vsi_seid != 0) {
10006                 dev_info(&pf->pdev->dev, "vsi seid %d not found\n",
10007                          vsi_seid);
10008                 return NULL;
10009         }
10010
10011         if (uplink_seid && uplink_seid != pf->mac_seid) {
10012                 for (veb_idx = 0; veb_idx < I40E_MAX_VEB; veb_idx++) {
10013                         if (pf->veb[veb_idx] &&
10014                             pf->veb[veb_idx]->seid == uplink_seid) {
10015                                 uplink_veb = pf->veb[veb_idx];
10016                                 break;
10017                         }
10018                 }
10019                 if (!uplink_veb) {
10020                         dev_info(&pf->pdev->dev,
10021                                  "uplink seid %d not found\n", uplink_seid);
10022                         return NULL;
10023                 }
10024         }
10025
10026         /* get veb sw struct */
10027         veb_idx = i40e_veb_mem_alloc(pf);
10028         if (veb_idx < 0)
10029                 goto err_alloc;
10030         veb = pf->veb[veb_idx];
10031         veb->flags = flags;
10032         veb->uplink_seid = uplink_seid;
10033         veb->veb_idx = (uplink_veb ? uplink_veb->idx : I40E_NO_VEB);
10034         veb->enabled_tc = (enabled_tc ? enabled_tc : 0x1);
10035
10036         /* create the VEB in the switch */
10037         ret = i40e_add_veb(veb, pf->vsi[vsi_idx]);
10038         if (ret)
10039                 goto err_veb;
10040         if (vsi_idx == pf->lan_vsi)
10041                 pf->lan_veb = veb->idx;
10042
10043         return veb;
10044
10045 err_veb:
10046         i40e_veb_clear(veb);
10047 err_alloc:
10048         return NULL;
10049 }
10050
10051 /**
10052  * i40e_setup_pf_switch_element - set PF vars based on switch type
10053  * @pf: board private structure
10054  * @ele: element we are building info from
10055  * @num_reported: total number of elements
10056  * @printconfig: should we print the contents
10057  *
10058  * helper function to assist in extracting a few useful SEID values.
10059  **/
10060 static void i40e_setup_pf_switch_element(struct i40e_pf *pf,
10061                                 struct i40e_aqc_switch_config_element_resp *ele,
10062                                 u16 num_reported, bool printconfig)
10063 {
10064         u16 downlink_seid = le16_to_cpu(ele->downlink_seid);
10065         u16 uplink_seid = le16_to_cpu(ele->uplink_seid);
10066         u8 element_type = ele->element_type;
10067         u16 seid = le16_to_cpu(ele->seid);
10068
10069         if (printconfig)
10070                 dev_info(&pf->pdev->dev,
10071                          "type=%d seid=%d uplink=%d downlink=%d\n",
10072                          element_type, seid, uplink_seid, downlink_seid);
10073
10074         switch (element_type) {
10075         case I40E_SWITCH_ELEMENT_TYPE_MAC:
10076                 pf->mac_seid = seid;
10077                 break;
10078         case I40E_SWITCH_ELEMENT_TYPE_VEB:
10079                 /* Main VEB? */
10080                 if (uplink_seid != pf->mac_seid)
10081                         break;
10082                 if (pf->lan_veb == I40E_NO_VEB) {
10083                         int v;
10084
10085                         /* find existing or else empty VEB */
10086                         for (v = 0; v < I40E_MAX_VEB; v++) {
10087                                 if (pf->veb[v] && (pf->veb[v]->seid == seid)) {
10088                                         pf->lan_veb = v;
10089                                         break;
10090                                 }
10091                         }
10092                         if (pf->lan_veb == I40E_NO_VEB) {
10093                                 v = i40e_veb_mem_alloc(pf);
10094                                 if (v < 0)
10095                                         break;
10096                                 pf->lan_veb = v;
10097                         }
10098                 }
10099
10100                 pf->veb[pf->lan_veb]->seid = seid;
10101                 pf->veb[pf->lan_veb]->uplink_seid = pf->mac_seid;
10102                 pf->veb[pf->lan_veb]->pf = pf;
10103                 pf->veb[pf->lan_veb]->veb_idx = I40E_NO_VEB;
10104                 break;
10105         case I40E_SWITCH_ELEMENT_TYPE_VSI:
10106                 if (num_reported != 1)
10107                         break;
10108                 /* This is immediately after a reset so we can assume this is
10109                  * the PF's VSI
10110                  */
10111                 pf->mac_seid = uplink_seid;
10112                 pf->pf_seid = downlink_seid;
10113                 pf->main_vsi_seid = seid;
10114                 if (printconfig)
10115                         dev_info(&pf->pdev->dev,
10116                                  "pf_seid=%d main_vsi_seid=%d\n",
10117                                  pf->pf_seid, pf->main_vsi_seid);
10118                 break;
10119         case I40E_SWITCH_ELEMENT_TYPE_PF:
10120         case I40E_SWITCH_ELEMENT_TYPE_VF:
10121         case I40E_SWITCH_ELEMENT_TYPE_EMP:
10122         case I40E_SWITCH_ELEMENT_TYPE_BMC:
10123         case I40E_SWITCH_ELEMENT_TYPE_PE:
10124         case I40E_SWITCH_ELEMENT_TYPE_PA:
10125                 /* ignore these for now */
10126                 break;
10127         default:
10128                 dev_info(&pf->pdev->dev, "unknown element type=%d seid=%d\n",
10129                          element_type, seid);
10130                 break;
10131         }
10132 }
10133
10134 /**
10135  * i40e_fetch_switch_configuration - Get switch config from firmware
10136  * @pf: board private structure
10137  * @printconfig: should we print the contents
10138  *
10139  * Get the current switch configuration from the device and
10140  * extract a few useful SEID values.
10141  **/
10142 int i40e_fetch_switch_configuration(struct i40e_pf *pf, bool printconfig)
10143 {
10144         struct i40e_aqc_get_switch_config_resp *sw_config;
10145         u16 next_seid = 0;
10146         int ret = 0;
10147         u8 *aq_buf;
10148         int i;
10149
10150         aq_buf = kzalloc(I40E_AQ_LARGE_BUF, GFP_KERNEL);
10151         if (!aq_buf)
10152                 return -ENOMEM;
10153
10154         sw_config = (struct i40e_aqc_get_switch_config_resp *)aq_buf;
10155         do {
10156                 u16 num_reported, num_total;
10157
10158                 ret = i40e_aq_get_switch_config(&pf->hw, sw_config,
10159                                                 I40E_AQ_LARGE_BUF,
10160                                                 &next_seid, NULL);
10161                 if (ret) {
10162                         dev_info(&pf->pdev->dev,
10163                                  "get switch config failed err %s aq_err %s\n",
10164                                  i40e_stat_str(&pf->hw, ret),
10165                                  i40e_aq_str(&pf->hw,
10166                                              pf->hw.aq.asq_last_status));
10167                         kfree(aq_buf);
10168                         return -ENOENT;
10169                 }
10170
10171                 num_reported = le16_to_cpu(sw_config->header.num_reported);
10172                 num_total = le16_to_cpu(sw_config->header.num_total);
10173
10174                 if (printconfig)
10175                         dev_info(&pf->pdev->dev,
10176                                  "header: %d reported %d total\n",
10177                                  num_reported, num_total);
10178
10179                 for (i = 0; i < num_reported; i++) {
10180                         struct i40e_aqc_switch_config_element_resp *ele =
10181                                 &sw_config->element[i];
10182
10183                         i40e_setup_pf_switch_element(pf, ele, num_reported,
10184                                                      printconfig);
10185                 }
10186         } while (next_seid != 0);
10187
10188         kfree(aq_buf);
10189         return ret;
10190 }
10191
10192 /**
10193  * i40e_setup_pf_switch - Setup the HW switch on startup or after reset
10194  * @pf: board private structure
10195  * @reinit: if the Main VSI needs to re-initialized.
10196  *
10197  * Returns 0 on success, negative value on failure
10198  **/
10199 static int i40e_setup_pf_switch(struct i40e_pf *pf, bool reinit)
10200 {
10201         int ret;
10202
10203         /* find out what's out there already */
10204         ret = i40e_fetch_switch_configuration(pf, false);
10205         if (ret) {
10206                 dev_info(&pf->pdev->dev,
10207                          "couldn't fetch switch config, err %s aq_err %s\n",
10208                          i40e_stat_str(&pf->hw, ret),
10209                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10210                 return ret;
10211         }
10212         i40e_pf_reset_stats(pf);
10213
10214         /* first time setup */
10215         if (pf->lan_vsi == I40E_NO_VSI || reinit) {
10216                 struct i40e_vsi *vsi = NULL;
10217                 u16 uplink_seid;
10218
10219                 /* Set up the PF VSI associated with the PF's main VSI
10220                  * that is already in the HW switch
10221                  */
10222                 if (pf->lan_veb != I40E_NO_VEB && pf->veb[pf->lan_veb])
10223                         uplink_seid = pf->veb[pf->lan_veb]->seid;
10224                 else
10225                         uplink_seid = pf->mac_seid;
10226                 if (pf->lan_vsi == I40E_NO_VSI)
10227                         vsi = i40e_vsi_setup(pf, I40E_VSI_MAIN, uplink_seid, 0);
10228                 else if (reinit)
10229                         vsi = i40e_vsi_reinit_setup(pf->vsi[pf->lan_vsi]);
10230                 if (!vsi) {
10231                         dev_info(&pf->pdev->dev, "setup of MAIN VSI failed\n");
10232                         i40e_fdir_teardown(pf);
10233                         return -EAGAIN;
10234                 }
10235         } else {
10236                 /* force a reset of TC and queue layout configurations */
10237                 u8 enabled_tc = pf->vsi[pf->lan_vsi]->tc_config.enabled_tc;
10238
10239                 pf->vsi[pf->lan_vsi]->tc_config.enabled_tc = 0;
10240                 pf->vsi[pf->lan_vsi]->seid = pf->main_vsi_seid;
10241                 i40e_vsi_config_tc(pf->vsi[pf->lan_vsi], enabled_tc);
10242         }
10243         i40e_vlan_stripping_disable(pf->vsi[pf->lan_vsi]);
10244
10245         i40e_fdir_sb_setup(pf);
10246
10247         /* Setup static PF queue filter control settings */
10248         ret = i40e_setup_pf_filter_control(pf);
10249         if (ret) {
10250                 dev_info(&pf->pdev->dev, "setup_pf_filter_control failed: %d\n",
10251                          ret);
10252                 /* Failure here should not stop continuing other steps */
10253         }
10254
10255         /* enable RSS in the HW, even for only one queue, as the stack can use
10256          * the hash
10257          */
10258         if ((pf->flags & I40E_FLAG_RSS_ENABLED))
10259                 i40e_pf_config_rss(pf);
10260
10261         /* fill in link information and enable LSE reporting */
10262         i40e_update_link_info(&pf->hw);
10263         i40e_link_event(pf);
10264
10265         /* Initialize user-specific link properties */
10266         pf->fc_autoneg_status = ((pf->hw.phy.link_info.an_info &
10267                                   I40E_AQ_AN_COMPLETED) ? true : false);
10268
10269         i40e_ptp_init(pf);
10270
10271         return ret;
10272 }
10273
10274 /**
10275  * i40e_determine_queue_usage - Work out queue distribution
10276  * @pf: board private structure
10277  **/
10278 static void i40e_determine_queue_usage(struct i40e_pf *pf)
10279 {
10280         int queues_left;
10281
10282         pf->num_lan_qps = 0;
10283 #ifdef I40E_FCOE
10284         pf->num_fcoe_qps = 0;
10285 #endif
10286
10287         /* Find the max queues to be put into basic use.  We'll always be
10288          * using TC0, whether or not DCB is running, and TC0 will get the
10289          * big RSS set.
10290          */
10291         queues_left = pf->hw.func_caps.num_tx_qp;
10292
10293         if ((queues_left == 1) ||
10294             !(pf->flags & I40E_FLAG_MSIX_ENABLED)) {
10295                 /* one qp for PF, no queues for anything else */
10296                 queues_left = 0;
10297                 pf->alloc_rss_size = pf->num_lan_qps = 1;
10298
10299                 /* make sure all the fancies are disabled */
10300                 pf->flags &= ~(I40E_FLAG_RSS_ENABLED    |
10301 #ifdef I40E_FCOE
10302                                I40E_FLAG_FCOE_ENABLED   |
10303 #endif
10304                                I40E_FLAG_FD_SB_ENABLED  |
10305                                I40E_FLAG_FD_ATR_ENABLED |
10306                                I40E_FLAG_DCB_CAPABLE    |
10307                                I40E_FLAG_SRIOV_ENABLED  |
10308                                I40E_FLAG_VMDQ_ENABLED);
10309         } else if (!(pf->flags & (I40E_FLAG_RSS_ENABLED |
10310                                   I40E_FLAG_FD_SB_ENABLED |
10311                                   I40E_FLAG_FD_ATR_ENABLED |
10312                                   I40E_FLAG_DCB_CAPABLE))) {
10313                 /* one qp for PF */
10314                 pf->alloc_rss_size = pf->num_lan_qps = 1;
10315                 queues_left -= pf->num_lan_qps;
10316
10317                 pf->flags &= ~(I40E_FLAG_RSS_ENABLED    |
10318 #ifdef I40E_FCOE
10319                                I40E_FLAG_FCOE_ENABLED   |
10320 #endif
10321                                I40E_FLAG_FD_SB_ENABLED  |
10322                                I40E_FLAG_FD_ATR_ENABLED |
10323                                I40E_FLAG_DCB_ENABLED    |
10324                                I40E_FLAG_VMDQ_ENABLED);
10325         } else {
10326                 /* Not enough queues for all TCs */
10327                 if ((pf->flags & I40E_FLAG_DCB_CAPABLE) &&
10328                     (queues_left < I40E_MAX_TRAFFIC_CLASS)) {
10329                         pf->flags &= ~I40E_FLAG_DCB_CAPABLE;
10330                         dev_info(&pf->pdev->dev, "not enough queues for DCB. DCB is disabled.\n");
10331                 }
10332                 pf->num_lan_qps = max_t(int, pf->rss_size_max,
10333                                         num_online_cpus());
10334                 pf->num_lan_qps = min_t(int, pf->num_lan_qps,
10335                                         pf->hw.func_caps.num_tx_qp);
10336
10337                 queues_left -= pf->num_lan_qps;
10338         }
10339
10340 #ifdef I40E_FCOE
10341         if (pf->flags & I40E_FLAG_FCOE_ENABLED) {
10342                 if (I40E_DEFAULT_FCOE <= queues_left) {
10343                         pf->num_fcoe_qps = I40E_DEFAULT_FCOE;
10344                 } else if (I40E_MINIMUM_FCOE <= queues_left) {
10345                         pf->num_fcoe_qps = I40E_MINIMUM_FCOE;
10346                 } else {
10347                         pf->num_fcoe_qps = 0;
10348                         pf->flags &= ~I40E_FLAG_FCOE_ENABLED;
10349                         dev_info(&pf->pdev->dev, "not enough queues for FCoE. FCoE feature will be disabled\n");
10350                 }
10351
10352                 queues_left -= pf->num_fcoe_qps;
10353         }
10354
10355 #endif
10356         if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
10357                 if (queues_left > 1) {
10358                         queues_left -= 1; /* save 1 queue for FD */
10359                 } else {
10360                         pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
10361                         dev_info(&pf->pdev->dev, "not enough queues for Flow Director. Flow Director feature is disabled\n");
10362                 }
10363         }
10364
10365         if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
10366             pf->num_vf_qps && pf->num_req_vfs && queues_left) {
10367                 pf->num_req_vfs = min_t(int, pf->num_req_vfs,
10368                                         (queues_left / pf->num_vf_qps));
10369                 queues_left -= (pf->num_req_vfs * pf->num_vf_qps);
10370         }
10371
10372         if ((pf->flags & I40E_FLAG_VMDQ_ENABLED) &&
10373             pf->num_vmdq_vsis && pf->num_vmdq_qps && queues_left) {
10374                 pf->num_vmdq_vsis = min_t(int, pf->num_vmdq_vsis,
10375                                           (queues_left / pf->num_vmdq_qps));
10376                 queues_left -= (pf->num_vmdq_vsis * pf->num_vmdq_qps);
10377         }
10378
10379         pf->queues_left = queues_left;
10380         dev_dbg(&pf->pdev->dev,
10381                 "qs_avail=%d FD SB=%d lan_qs=%d lan_tc0=%d vf=%d*%d vmdq=%d*%d, remaining=%d\n",
10382                 pf->hw.func_caps.num_tx_qp,
10383                 !!(pf->flags & I40E_FLAG_FD_SB_ENABLED),
10384                 pf->num_lan_qps, pf->alloc_rss_size, pf->num_req_vfs,
10385                 pf->num_vf_qps, pf->num_vmdq_vsis, pf->num_vmdq_qps,
10386                 queues_left);
10387 #ifdef I40E_FCOE
10388         dev_dbg(&pf->pdev->dev, "fcoe queues = %d\n", pf->num_fcoe_qps);
10389 #endif
10390 }
10391
10392 /**
10393  * i40e_setup_pf_filter_control - Setup PF static filter control
10394  * @pf: PF to be setup
10395  *
10396  * i40e_setup_pf_filter_control sets up a PF's initial filter control
10397  * settings. If PE/FCoE are enabled then it will also set the per PF
10398  * based filter sizes required for them. It also enables Flow director,
10399  * ethertype and macvlan type filter settings for the pf.
10400  *
10401  * Returns 0 on success, negative on failure
10402  **/
10403 static int i40e_setup_pf_filter_control(struct i40e_pf *pf)
10404 {
10405         struct i40e_filter_control_settings *settings = &pf->filter_settings;
10406
10407         settings->hash_lut_size = I40E_HASH_LUT_SIZE_128;
10408
10409         /* Flow Director is enabled */
10410         if (pf->flags & (I40E_FLAG_FD_SB_ENABLED | I40E_FLAG_FD_ATR_ENABLED))
10411                 settings->enable_fdir = true;
10412
10413         /* Ethtype and MACVLAN filters enabled for PF */
10414         settings->enable_ethtype = true;
10415         settings->enable_macvlan = true;
10416
10417         if (i40e_set_filter_control(&pf->hw, settings))
10418                 return -ENOENT;
10419
10420         return 0;
10421 }
10422
10423 #define INFO_STRING_LEN 255
10424 #define REMAIN(__x) (INFO_STRING_LEN - (__x))
10425 static void i40e_print_features(struct i40e_pf *pf)
10426 {
10427         struct i40e_hw *hw = &pf->hw;
10428         char *buf;
10429         int i;
10430
10431         buf = kmalloc(INFO_STRING_LEN, GFP_KERNEL);
10432         if (!buf)
10433                 return;
10434
10435         i = snprintf(buf, INFO_STRING_LEN, "Features: PF-id[%d]", hw->pf_id);
10436 #ifdef CONFIG_PCI_IOV
10437         i += snprintf(&buf[i], REMAIN(i), " VFs: %d", pf->num_req_vfs);
10438 #endif
10439         i += snprintf(&buf[i], REMAIN(i), " VSIs: %d QP: %d RX: %s",
10440                       pf->hw.func_caps.num_vsis,
10441                       pf->vsi[pf->lan_vsi]->num_queue_pairs,
10442                       pf->flags & I40E_FLAG_RX_PS_ENABLED ? "PS" : "1BUF");
10443
10444         if (pf->flags & I40E_FLAG_RSS_ENABLED)
10445                 i += snprintf(&buf[i], REMAIN(i), " RSS");
10446         if (pf->flags & I40E_FLAG_FD_ATR_ENABLED)
10447                 i += snprintf(&buf[i], REMAIN(i), " FD_ATR");
10448         if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
10449                 i += snprintf(&buf[i], REMAIN(i), " FD_SB");
10450                 i += snprintf(&buf[i], REMAIN(i), " NTUPLE");
10451         }
10452         if (pf->flags & I40E_FLAG_DCB_CAPABLE)
10453                 i += snprintf(&buf[i], REMAIN(i), " DCB");
10454 #if IS_ENABLED(CONFIG_VXLAN)
10455         i += snprintf(&buf[i], REMAIN(i), " VxLAN");
10456 #endif
10457 #if IS_ENABLED(CONFIG_GENEVE)
10458         i += snprintf(&buf[i], REMAIN(i), " Geneve");
10459 #endif
10460         if (pf->flags & I40E_FLAG_PTP)
10461                 i += snprintf(&buf[i], REMAIN(i), " PTP");
10462 #ifdef I40E_FCOE
10463         if (pf->flags & I40E_FLAG_FCOE_ENABLED)
10464                 i += snprintf(&buf[i], REMAIN(i), " FCOE");
10465 #endif
10466         if (pf->flags & I40E_FLAG_VEB_MODE_ENABLED)
10467                 i += snprintf(&buf[i], REMAIN(i), " VEB");
10468         else
10469                 i += snprintf(&buf[i], REMAIN(i), " VEPA");
10470
10471         dev_info(&pf->pdev->dev, "%s\n", buf);
10472         kfree(buf);
10473         WARN_ON(i > INFO_STRING_LEN);
10474 }
10475
10476 /**
10477  * i40e_probe - Device initialization routine
10478  * @pdev: PCI device information struct
10479  * @ent: entry in i40e_pci_tbl
10480  *
10481  * i40e_probe initializes a PF identified by a pci_dev structure.
10482  * The OS initialization, configuring of the PF private structure,
10483  * and a hardware reset occur.
10484  *
10485  * Returns 0 on success, negative on failure
10486  **/
10487 static int i40e_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
10488 {
10489         struct i40e_aq_get_phy_abilities_resp abilities;
10490         struct i40e_pf *pf;
10491         struct i40e_hw *hw;
10492         static u16 pfs_found;
10493         u16 wol_nvm_bits;
10494         u16 link_status;
10495         int err;
10496         u32 len;
10497         u32 val;
10498         u32 i;
10499         u8 set_fc_aq_fail;
10500
10501         err = pci_enable_device_mem(pdev);
10502         if (err)
10503                 return err;
10504
10505         /* set up for high or low dma */
10506         err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
10507         if (err) {
10508                 err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
10509                 if (err) {
10510                         dev_err(&pdev->dev,
10511                                 "DMA configuration failed: 0x%x\n", err);
10512                         goto err_dma;
10513                 }
10514         }
10515
10516         /* set up pci connections */
10517         err = pci_request_selected_regions(pdev, pci_select_bars(pdev,
10518                                            IORESOURCE_MEM), i40e_driver_name);
10519         if (err) {
10520                 dev_info(&pdev->dev,
10521                          "pci_request_selected_regions failed %d\n", err);
10522                 goto err_pci_reg;
10523         }
10524
10525         pci_enable_pcie_error_reporting(pdev);
10526         pci_set_master(pdev);
10527
10528         /* Now that we have a PCI connection, we need to do the
10529          * low level device setup.  This is primarily setting up
10530          * the Admin Queue structures and then querying for the
10531          * device's current profile information.
10532          */
10533         pf = kzalloc(sizeof(*pf), GFP_KERNEL);
10534         if (!pf) {
10535                 err = -ENOMEM;
10536                 goto err_pf_alloc;
10537         }
10538         pf->next_vsi = 0;
10539         pf->pdev = pdev;
10540         set_bit(__I40E_DOWN, &pf->state);
10541
10542         hw = &pf->hw;
10543         hw->back = pf;
10544
10545         pf->ioremap_len = min_t(int, pci_resource_len(pdev, 0),
10546                                 I40E_MAX_CSR_SPACE);
10547
10548         hw->hw_addr = ioremap(pci_resource_start(pdev, 0), pf->ioremap_len);
10549         if (!hw->hw_addr) {
10550                 err = -EIO;
10551                 dev_info(&pdev->dev, "ioremap(0x%04x, 0x%04x) failed: 0x%x\n",
10552                          (unsigned int)pci_resource_start(pdev, 0),
10553                          pf->ioremap_len, err);
10554                 goto err_ioremap;
10555         }
10556         hw->vendor_id = pdev->vendor;
10557         hw->device_id = pdev->device;
10558         pci_read_config_byte(pdev, PCI_REVISION_ID, &hw->revision_id);
10559         hw->subsystem_vendor_id = pdev->subsystem_vendor;
10560         hw->subsystem_device_id = pdev->subsystem_device;
10561         hw->bus.device = PCI_SLOT(pdev->devfn);
10562         hw->bus.func = PCI_FUNC(pdev->devfn);
10563         pf->instance = pfs_found;
10564
10565         if (debug != -1) {
10566                 pf->msg_enable = pf->hw.debug_mask;
10567                 pf->msg_enable = debug;
10568         }
10569
10570         /* do a special CORER for clearing PXE mode once at init */
10571         if (hw->revision_id == 0 &&
10572             (rd32(hw, I40E_GLLAN_RCTL_0) & I40E_GLLAN_RCTL_0_PXE_MODE_MASK)) {
10573                 wr32(hw, I40E_GLGEN_RTRIG, I40E_GLGEN_RTRIG_CORER_MASK);
10574                 i40e_flush(hw);
10575                 msleep(200);
10576                 pf->corer_count++;
10577
10578                 i40e_clear_pxe_mode(hw);
10579         }
10580
10581         /* Reset here to make sure all is clean and to define PF 'n' */
10582         i40e_clear_hw(hw);
10583         err = i40e_pf_reset(hw);
10584         if (err) {
10585                 dev_info(&pdev->dev, "Initial pf_reset failed: %d\n", err);
10586                 goto err_pf_reset;
10587         }
10588         pf->pfr_count++;
10589
10590         hw->aq.num_arq_entries = I40E_AQ_LEN;
10591         hw->aq.num_asq_entries = I40E_AQ_LEN;
10592         hw->aq.arq_buf_size = I40E_MAX_AQ_BUF_SIZE;
10593         hw->aq.asq_buf_size = I40E_MAX_AQ_BUF_SIZE;
10594         pf->adminq_work_limit = I40E_AQ_WORK_LIMIT;
10595
10596         snprintf(pf->int_name, sizeof(pf->int_name) - 1,
10597                  "%s-%s:misc",
10598                  dev_driver_string(&pf->pdev->dev), dev_name(&pdev->dev));
10599
10600         err = i40e_init_shared_code(hw);
10601         if (err) {
10602                 dev_warn(&pdev->dev, "unidentified MAC or BLANK NVM: %d\n",
10603                          err);
10604                 goto err_pf_reset;
10605         }
10606
10607         /* set up a default setting for link flow control */
10608         pf->hw.fc.requested_mode = I40E_FC_NONE;
10609
10610         /* set up the locks for the AQ, do this only once in probe
10611          * and destroy them only once in remove
10612          */
10613         mutex_init(&hw->aq.asq_mutex);
10614         mutex_init(&hw->aq.arq_mutex);
10615
10616         err = i40e_init_adminq(hw);
10617         if (err) {
10618                 if (err == I40E_ERR_FIRMWARE_API_VERSION)
10619                         dev_info(&pdev->dev,
10620                                  "The driver for the device stopped because the NVM image is newer than expected. You must install the most recent version of the network driver.\n");
10621                 else
10622                         dev_info(&pdev->dev,
10623                                  "The driver for the device stopped because the device firmware failed to init. Try updating your NVM image.\n");
10624
10625                 goto err_pf_reset;
10626         }
10627
10628         /* provide nvm, fw, api versions */
10629         dev_info(&pdev->dev, "fw %d.%d.%05d api %d.%d nvm %s\n",
10630                  hw->aq.fw_maj_ver, hw->aq.fw_min_ver, hw->aq.fw_build,
10631                  hw->aq.api_maj_ver, hw->aq.api_min_ver,
10632                  i40e_nvm_version_str(hw));
10633
10634         if (hw->aq.api_maj_ver == I40E_FW_API_VERSION_MAJOR &&
10635             hw->aq.api_min_ver > I40E_FW_API_VERSION_MINOR)
10636                 dev_info(&pdev->dev,
10637                          "The driver for the device detected a newer version of the NVM image than expected. Please install the most recent version of the network driver.\n");
10638         else if (hw->aq.api_maj_ver < I40E_FW_API_VERSION_MAJOR ||
10639                  hw->aq.api_min_ver < (I40E_FW_API_VERSION_MINOR - 1))
10640                 dev_info(&pdev->dev,
10641                          "The driver for the device detected an older version of the NVM image than expected. Please update the NVM image.\n");
10642
10643         i40e_verify_eeprom(pf);
10644
10645         /* Rev 0 hardware was never productized */
10646         if (hw->revision_id < 1)
10647                 dev_warn(&pdev->dev, "This device is a pre-production adapter/LOM. Please be aware there may be issues with your hardware. If you are experiencing problems please contact your Intel or hardware representative who provided you with this hardware.\n");
10648
10649         i40e_clear_pxe_mode(hw);
10650         err = i40e_get_capabilities(pf);
10651         if (err)
10652                 goto err_adminq_setup;
10653
10654         err = i40e_sw_init(pf);
10655         if (err) {
10656                 dev_info(&pdev->dev, "sw_init failed: %d\n", err);
10657                 goto err_sw_init;
10658         }
10659
10660         err = i40e_init_lan_hmc(hw, hw->func_caps.num_tx_qp,
10661                                 hw->func_caps.num_rx_qp,
10662                                 pf->fcoe_hmc_cntx_num, pf->fcoe_hmc_filt_num);
10663         if (err) {
10664                 dev_info(&pdev->dev, "init_lan_hmc failed: %d\n", err);
10665                 goto err_init_lan_hmc;
10666         }
10667
10668         err = i40e_configure_lan_hmc(hw, I40E_HMC_MODEL_DIRECT_ONLY);
10669         if (err) {
10670                 dev_info(&pdev->dev, "configure_lan_hmc failed: %d\n", err);
10671                 err = -ENOENT;
10672                 goto err_configure_lan_hmc;
10673         }
10674
10675         /* Disable LLDP for NICs that have firmware versions lower than v4.3.
10676          * Ignore error return codes because if it was already disabled via
10677          * hardware settings this will fail
10678          */
10679         if (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver < 3)) ||
10680             (pf->hw.aq.fw_maj_ver < 4)) {
10681                 dev_info(&pdev->dev, "Stopping firmware LLDP agent.\n");
10682                 i40e_aq_stop_lldp(hw, true, NULL);
10683         }
10684
10685         i40e_get_mac_addr(hw, hw->mac.addr);
10686         if (!is_valid_ether_addr(hw->mac.addr)) {
10687                 dev_info(&pdev->dev, "invalid MAC address %pM\n", hw->mac.addr);
10688                 err = -EIO;
10689                 goto err_mac_addr;
10690         }
10691         dev_info(&pdev->dev, "MAC address: %pM\n", hw->mac.addr);
10692         ether_addr_copy(hw->mac.perm_addr, hw->mac.addr);
10693         i40e_get_port_mac_addr(hw, hw->mac.port_addr);
10694         if (is_valid_ether_addr(hw->mac.port_addr))
10695                 pf->flags |= I40E_FLAG_PORT_ID_VALID;
10696 #ifdef I40E_FCOE
10697         err = i40e_get_san_mac_addr(hw, hw->mac.san_addr);
10698         if (err)
10699                 dev_info(&pdev->dev,
10700                          "(non-fatal) SAN MAC retrieval failed: %d\n", err);
10701         if (!is_valid_ether_addr(hw->mac.san_addr)) {
10702                 dev_warn(&pdev->dev, "invalid SAN MAC address %pM, falling back to LAN MAC\n",
10703                          hw->mac.san_addr);
10704                 ether_addr_copy(hw->mac.san_addr, hw->mac.addr);
10705         }
10706         dev_info(&pf->pdev->dev, "SAN MAC: %pM\n", hw->mac.san_addr);
10707 #endif /* I40E_FCOE */
10708
10709         pci_set_drvdata(pdev, pf);
10710         pci_save_state(pdev);
10711 #ifdef CONFIG_I40E_DCB
10712         err = i40e_init_pf_dcb(pf);
10713         if (err) {
10714                 dev_info(&pdev->dev, "DCB init failed %d, disabled\n", err);
10715                 pf->flags &= ~I40E_FLAG_DCB_CAPABLE;
10716                 /* Continue without DCB enabled */
10717         }
10718 #endif /* CONFIG_I40E_DCB */
10719
10720         /* set up periodic task facility */
10721         setup_timer(&pf->service_timer, i40e_service_timer, (unsigned long)pf);
10722         pf->service_timer_period = HZ;
10723
10724         INIT_WORK(&pf->service_task, i40e_service_task);
10725         clear_bit(__I40E_SERVICE_SCHED, &pf->state);
10726         pf->flags |= I40E_FLAG_NEED_LINK_UPDATE;
10727
10728         /* NVM bit on means WoL disabled for the port */
10729         i40e_read_nvm_word(hw, I40E_SR_NVM_WAKE_ON_LAN, &wol_nvm_bits);
10730         if (BIT (hw->port) & wol_nvm_bits || hw->partition_id != 1)
10731                 pf->wol_en = false;
10732         else
10733                 pf->wol_en = true;
10734         device_set_wakeup_enable(&pf->pdev->dev, pf->wol_en);
10735
10736         /* set up the main switch operations */
10737         i40e_determine_queue_usage(pf);
10738         err = i40e_init_interrupt_scheme(pf);
10739         if (err)
10740                 goto err_switch_setup;
10741
10742         /* The number of VSIs reported by the FW is the minimum guaranteed
10743          * to us; HW supports far more and we share the remaining pool with
10744          * the other PFs. We allocate space for more than the guarantee with
10745          * the understanding that we might not get them all later.
10746          */
10747         if (pf->hw.func_caps.num_vsis < I40E_MIN_VSI_ALLOC)
10748                 pf->num_alloc_vsi = I40E_MIN_VSI_ALLOC;
10749         else
10750                 pf->num_alloc_vsi = pf->hw.func_caps.num_vsis;
10751
10752         /* Set up the *vsi struct and our local tracking of the MAIN PF vsi. */
10753         len = sizeof(struct i40e_vsi *) * pf->num_alloc_vsi;
10754         pf->vsi = kzalloc(len, GFP_KERNEL);
10755         if (!pf->vsi) {
10756                 err = -ENOMEM;
10757                 goto err_switch_setup;
10758         }
10759
10760 #ifdef CONFIG_PCI_IOV
10761         /* prep for VF support */
10762         if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
10763             (pf->flags & I40E_FLAG_MSIX_ENABLED) &&
10764             !test_bit(__I40E_BAD_EEPROM, &pf->state)) {
10765                 if (pci_num_vf(pdev))
10766                         pf->flags |= I40E_FLAG_VEB_MODE_ENABLED;
10767         }
10768 #endif
10769         err = i40e_setup_pf_switch(pf, false);
10770         if (err) {
10771                 dev_info(&pdev->dev, "setup_pf_switch failed: %d\n", err);
10772                 goto err_vsis;
10773         }
10774
10775         /* Make sure flow control is set according to current settings */
10776         err = i40e_set_fc(hw, &set_fc_aq_fail, true);
10777         if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_GET)
10778                 dev_dbg(&pf->pdev->dev,
10779                         "Set fc with err %s aq_err %s on get_phy_cap\n",
10780                         i40e_stat_str(hw, err),
10781                         i40e_aq_str(hw, hw->aq.asq_last_status));
10782         if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_SET)
10783                 dev_dbg(&pf->pdev->dev,
10784                         "Set fc with err %s aq_err %s on set_phy_config\n",
10785                         i40e_stat_str(hw, err),
10786                         i40e_aq_str(hw, hw->aq.asq_last_status));
10787         if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_UPDATE)
10788                 dev_dbg(&pf->pdev->dev,
10789                         "Set fc with err %s aq_err %s on get_link_info\n",
10790                         i40e_stat_str(hw, err),
10791                         i40e_aq_str(hw, hw->aq.asq_last_status));
10792
10793         /* if FDIR VSI was set up, start it now */
10794         for (i = 0; i < pf->num_alloc_vsi; i++) {
10795                 if (pf->vsi[i] && pf->vsi[i]->type == I40E_VSI_FDIR) {
10796                         i40e_vsi_open(pf->vsi[i]);
10797                         break;
10798                 }
10799         }
10800
10801         /* driver is only interested in link up/down and module qualification
10802          * reports from firmware
10803          */
10804         err = i40e_aq_set_phy_int_mask(&pf->hw,
10805                                        I40E_AQ_EVENT_LINK_UPDOWN |
10806                                        I40E_AQ_EVENT_MODULE_QUAL_FAIL, NULL);
10807         if (err)
10808                 dev_info(&pf->pdev->dev, "set phy mask fail, err %s aq_err %s\n",
10809                          i40e_stat_str(&pf->hw, err),
10810                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10811
10812         /* Reconfigure hardware for allowing smaller MSS in the case
10813          * of TSO, so that we avoid the MDD being fired and causing
10814          * a reset in the case of small MSS+TSO.
10815          */
10816         val = rd32(hw, I40E_REG_MSS);
10817         if ((val & I40E_REG_MSS_MIN_MASK) > I40E_64BYTE_MSS) {
10818                 val &= ~I40E_REG_MSS_MIN_MASK;
10819                 val |= I40E_64BYTE_MSS;
10820                 wr32(hw, I40E_REG_MSS, val);
10821         }
10822
10823         if (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver < 33)) ||
10824             (pf->hw.aq.fw_maj_ver < 4)) {
10825                 msleep(75);
10826                 err = i40e_aq_set_link_restart_an(&pf->hw, true, NULL);
10827                 if (err)
10828                         dev_info(&pf->pdev->dev, "link restart failed, err %s aq_err %s\n",
10829                                  i40e_stat_str(&pf->hw, err),
10830                                  i40e_aq_str(&pf->hw,
10831                                              pf->hw.aq.asq_last_status));
10832         }
10833         /* The main driver is (mostly) up and happy. We need to set this state
10834          * before setting up the misc vector or we get a race and the vector
10835          * ends up disabled forever.
10836          */
10837         clear_bit(__I40E_DOWN, &pf->state);
10838
10839         /* In case of MSIX we are going to setup the misc vector right here
10840          * to handle admin queue events etc. In case of legacy and MSI
10841          * the misc functionality and queue processing is combined in
10842          * the same vector and that gets setup at open.
10843          */
10844         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
10845                 err = i40e_setup_misc_vector(pf);
10846                 if (err) {
10847                         dev_info(&pdev->dev,
10848                                  "setup of misc vector failed: %d\n", err);
10849                         goto err_vsis;
10850                 }
10851         }
10852
10853 #ifdef CONFIG_PCI_IOV
10854         /* prep for VF support */
10855         if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
10856             (pf->flags & I40E_FLAG_MSIX_ENABLED) &&
10857             !test_bit(__I40E_BAD_EEPROM, &pf->state)) {
10858                 u32 val;
10859
10860                 /* disable link interrupts for VFs */
10861                 val = rd32(hw, I40E_PFGEN_PORTMDIO_NUM);
10862                 val &= ~I40E_PFGEN_PORTMDIO_NUM_VFLINK_STAT_ENA_MASK;
10863                 wr32(hw, I40E_PFGEN_PORTMDIO_NUM, val);
10864                 i40e_flush(hw);
10865
10866                 if (pci_num_vf(pdev)) {
10867                         dev_info(&pdev->dev,
10868                                  "Active VFs found, allocating resources.\n");
10869                         err = i40e_alloc_vfs(pf, pci_num_vf(pdev));
10870                         if (err)
10871                                 dev_info(&pdev->dev,
10872                                          "Error %d allocating resources for existing VFs\n",
10873                                          err);
10874                 }
10875         }
10876 #endif /* CONFIG_PCI_IOV */
10877
10878         pfs_found++;
10879
10880         i40e_dbg_pf_init(pf);
10881
10882         /* tell the firmware that we're starting */
10883         i40e_send_version(pf);
10884
10885         /* since everything's happy, start the service_task timer */
10886         mod_timer(&pf->service_timer,
10887                   round_jiffies(jiffies + pf->service_timer_period));
10888
10889 #ifdef I40E_FCOE
10890         /* create FCoE interface */
10891         i40e_fcoe_vsi_setup(pf);
10892
10893 #endif
10894 #define PCI_SPEED_SIZE 8
10895 #define PCI_WIDTH_SIZE 8
10896         /* Devices on the IOSF bus do not have this information
10897          * and will report PCI Gen 1 x 1 by default so don't bother
10898          * checking them.
10899          */
10900         if (!(pf->flags & I40E_FLAG_NO_PCI_LINK_CHECK)) {
10901                 char speed[PCI_SPEED_SIZE] = "Unknown";
10902                 char width[PCI_WIDTH_SIZE] = "Unknown";
10903
10904                 /* Get the negotiated link width and speed from PCI config
10905                  * space
10906                  */
10907                 pcie_capability_read_word(pf->pdev, PCI_EXP_LNKSTA,
10908                                           &link_status);
10909
10910                 i40e_set_pci_config_data(hw, link_status);
10911
10912                 switch (hw->bus.speed) {
10913                 case i40e_bus_speed_8000:
10914                         strncpy(speed, "8.0", PCI_SPEED_SIZE); break;
10915                 case i40e_bus_speed_5000:
10916                         strncpy(speed, "5.0", PCI_SPEED_SIZE); break;
10917                 case i40e_bus_speed_2500:
10918                         strncpy(speed, "2.5", PCI_SPEED_SIZE); break;
10919                 default:
10920                         break;
10921                 }
10922                 switch (hw->bus.width) {
10923                 case i40e_bus_width_pcie_x8:
10924                         strncpy(width, "8", PCI_WIDTH_SIZE); break;
10925                 case i40e_bus_width_pcie_x4:
10926                         strncpy(width, "4", PCI_WIDTH_SIZE); break;
10927                 case i40e_bus_width_pcie_x2:
10928                         strncpy(width, "2", PCI_WIDTH_SIZE); break;
10929                 case i40e_bus_width_pcie_x1:
10930                         strncpy(width, "1", PCI_WIDTH_SIZE); break;
10931                 default:
10932                         break;
10933                 }
10934
10935                 dev_info(&pdev->dev, "PCI-Express: Speed %sGT/s Width x%s\n",
10936                          speed, width);
10937
10938                 if (hw->bus.width < i40e_bus_width_pcie_x8 ||
10939                     hw->bus.speed < i40e_bus_speed_8000) {
10940                         dev_warn(&pdev->dev, "PCI-Express bandwidth available for this device may be insufficient for optimal performance.\n");
10941                         dev_warn(&pdev->dev, "Please move the device to a different PCI-e link with more lanes and/or higher transfer rate.\n");
10942                 }
10943         }
10944
10945         /* get the requested speeds from the fw */
10946         err = i40e_aq_get_phy_capabilities(hw, false, false, &abilities, NULL);
10947         if (err)
10948                 dev_dbg(&pf->pdev->dev, "get requested speeds ret =  %s last_status =  %s\n",
10949                         i40e_stat_str(&pf->hw, err),
10950                         i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10951         pf->hw.phy.link_info.requested_speeds = abilities.link_speed;
10952
10953         /* get the supported phy types from the fw */
10954         err = i40e_aq_get_phy_capabilities(hw, false, true, &abilities, NULL);
10955         if (err)
10956                 dev_dbg(&pf->pdev->dev, "get supported phy types ret =  %s last_status =  %s\n",
10957                         i40e_stat_str(&pf->hw, err),
10958                         i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10959         pf->hw.phy.phy_types = le32_to_cpu(abilities.phy_type);
10960
10961         /* Add a filter to drop all Flow control frames from any VSI from being
10962          * transmitted. By doing so we stop a malicious VF from sending out
10963          * PAUSE or PFC frames and potentially controlling traffic for other
10964          * PF/VF VSIs.
10965          * The FW can still send Flow control frames if enabled.
10966          */
10967         i40e_add_filter_to_drop_tx_flow_control_frames(&pf->hw,
10968                                                        pf->main_vsi_seid);
10969
10970         /* print a string summarizing features */
10971         i40e_print_features(pf);
10972
10973         return 0;
10974
10975         /* Unwind what we've done if something failed in the setup */
10976 err_vsis:
10977         set_bit(__I40E_DOWN, &pf->state);
10978         i40e_clear_interrupt_scheme(pf);
10979         kfree(pf->vsi);
10980 err_switch_setup:
10981         i40e_reset_interrupt_capability(pf);
10982         del_timer_sync(&pf->service_timer);
10983 err_mac_addr:
10984 err_configure_lan_hmc:
10985         (void)i40e_shutdown_lan_hmc(hw);
10986 err_init_lan_hmc:
10987         kfree(pf->qp_pile);
10988 err_sw_init:
10989 err_adminq_setup:
10990         (void)i40e_shutdown_adminq(hw);
10991 err_pf_reset:
10992         iounmap(hw->hw_addr);
10993 err_ioremap:
10994         kfree(pf);
10995 err_pf_alloc:
10996         pci_disable_pcie_error_reporting(pdev);
10997         pci_release_selected_regions(pdev,
10998                                      pci_select_bars(pdev, IORESOURCE_MEM));
10999 err_pci_reg:
11000 err_dma:
11001         pci_disable_device(pdev);
11002         return err;
11003 }
11004
11005 /**
11006  * i40e_remove - Device removal routine
11007  * @pdev: PCI device information struct
11008  *
11009  * i40e_remove is called by the PCI subsystem to alert the driver
11010  * that is should release a PCI device.  This could be caused by a
11011  * Hot-Plug event, or because the driver is going to be removed from
11012  * memory.
11013  **/
11014 static void i40e_remove(struct pci_dev *pdev)
11015 {
11016         struct i40e_pf *pf = pci_get_drvdata(pdev);
11017         struct i40e_hw *hw = &pf->hw;
11018         i40e_status ret_code;
11019         int i;
11020
11021         i40e_dbg_pf_exit(pf);
11022
11023         i40e_ptp_stop(pf);
11024
11025         /* Disable RSS in hw */
11026         wr32(hw, I40E_PFQF_HENA(0), 0);
11027         wr32(hw, I40E_PFQF_HENA(1), 0);
11028
11029         /* no more scheduling of any task */
11030         set_bit(__I40E_DOWN, &pf->state);
11031         del_timer_sync(&pf->service_timer);
11032         cancel_work_sync(&pf->service_task);
11033
11034         if (pf->flags & I40E_FLAG_SRIOV_ENABLED) {
11035                 i40e_free_vfs(pf);
11036                 pf->flags &= ~I40E_FLAG_SRIOV_ENABLED;
11037         }
11038
11039         i40e_fdir_teardown(pf);
11040
11041         /* If there is a switch structure or any orphans, remove them.
11042          * This will leave only the PF's VSI remaining.
11043          */
11044         for (i = 0; i < I40E_MAX_VEB; i++) {
11045                 if (!pf->veb[i])
11046                         continue;
11047
11048                 if (pf->veb[i]->uplink_seid == pf->mac_seid ||
11049                     pf->veb[i]->uplink_seid == 0)
11050                         i40e_switch_branch_release(pf->veb[i]);
11051         }
11052
11053         /* Now we can shutdown the PF's VSI, just before we kill
11054          * adminq and hmc.
11055          */
11056         if (pf->vsi[pf->lan_vsi])
11057                 i40e_vsi_release(pf->vsi[pf->lan_vsi]);
11058
11059         /* shutdown and destroy the HMC */
11060         if (pf->hw.hmc.hmc_obj) {
11061                 ret_code = i40e_shutdown_lan_hmc(&pf->hw);
11062                 if (ret_code)
11063                         dev_warn(&pdev->dev,
11064                                  "Failed to destroy the HMC resources: %d\n",
11065                                  ret_code);
11066         }
11067
11068         /* shutdown the adminq */
11069         ret_code = i40e_shutdown_adminq(&pf->hw);
11070         if (ret_code)
11071                 dev_warn(&pdev->dev,
11072                          "Failed to destroy the Admin Queue resources: %d\n",
11073                          ret_code);
11074
11075         /* destroy the locks only once, here */
11076         mutex_destroy(&hw->aq.arq_mutex);
11077         mutex_destroy(&hw->aq.asq_mutex);
11078
11079         /* Clear all dynamic memory lists of rings, q_vectors, and VSIs */
11080         i40e_clear_interrupt_scheme(pf);
11081         for (i = 0; i < pf->num_alloc_vsi; i++) {
11082                 if (pf->vsi[i]) {
11083                         i40e_vsi_clear_rings(pf->vsi[i]);
11084                         i40e_vsi_clear(pf->vsi[i]);
11085                         pf->vsi[i] = NULL;
11086                 }
11087         }
11088
11089         for (i = 0; i < I40E_MAX_VEB; i++) {
11090                 kfree(pf->veb[i]);
11091                 pf->veb[i] = NULL;
11092         }
11093
11094         kfree(pf->qp_pile);
11095         kfree(pf->vsi);
11096
11097         iounmap(pf->hw.hw_addr);
11098         kfree(pf);
11099         pci_release_selected_regions(pdev,
11100                                      pci_select_bars(pdev, IORESOURCE_MEM));
11101
11102         pci_disable_pcie_error_reporting(pdev);
11103         pci_disable_device(pdev);
11104 }
11105
11106 /**
11107  * i40e_pci_error_detected - warning that something funky happened in PCI land
11108  * @pdev: PCI device information struct
11109  *
11110  * Called to warn that something happened and the error handling steps
11111  * are in progress.  Allows the driver to quiesce things, be ready for
11112  * remediation.
11113  **/
11114 static pci_ers_result_t i40e_pci_error_detected(struct pci_dev *pdev,
11115                                                 enum pci_channel_state error)
11116 {
11117         struct i40e_pf *pf = pci_get_drvdata(pdev);
11118
11119         dev_info(&pdev->dev, "%s: error %d\n", __func__, error);
11120
11121         /* shutdown all operations */
11122         if (!test_bit(__I40E_SUSPENDED, &pf->state)) {
11123                 rtnl_lock();
11124                 i40e_prep_for_reset(pf);
11125                 rtnl_unlock();
11126         }
11127
11128         /* Request a slot reset */
11129         return PCI_ERS_RESULT_NEED_RESET;
11130 }
11131
11132 /**
11133  * i40e_pci_error_slot_reset - a PCI slot reset just happened
11134  * @pdev: PCI device information struct
11135  *
11136  * Called to find if the driver can work with the device now that
11137  * the pci slot has been reset.  If a basic connection seems good
11138  * (registers are readable and have sane content) then return a
11139  * happy little PCI_ERS_RESULT_xxx.
11140  **/
11141 static pci_ers_result_t i40e_pci_error_slot_reset(struct pci_dev *pdev)
11142 {
11143         struct i40e_pf *pf = pci_get_drvdata(pdev);
11144         pci_ers_result_t result;
11145         int err;
11146         u32 reg;
11147
11148         dev_dbg(&pdev->dev, "%s\n", __func__);
11149         if (pci_enable_device_mem(pdev)) {
11150                 dev_info(&pdev->dev,
11151                          "Cannot re-enable PCI device after reset.\n");
11152                 result = PCI_ERS_RESULT_DISCONNECT;
11153         } else {
11154                 pci_set_master(pdev);
11155                 pci_restore_state(pdev);
11156                 pci_save_state(pdev);
11157                 pci_wake_from_d3(pdev, false);
11158
11159                 reg = rd32(&pf->hw, I40E_GLGEN_RTRIG);
11160                 if (reg == 0)
11161                         result = PCI_ERS_RESULT_RECOVERED;
11162                 else
11163                         result = PCI_ERS_RESULT_DISCONNECT;
11164         }
11165
11166         err = pci_cleanup_aer_uncorrect_error_status(pdev);
11167         if (err) {
11168                 dev_info(&pdev->dev,
11169                          "pci_cleanup_aer_uncorrect_error_status failed 0x%0x\n",
11170                          err);
11171                 /* non-fatal, continue */
11172         }
11173
11174         return result;
11175 }
11176
11177 /**
11178  * i40e_pci_error_resume - restart operations after PCI error recovery
11179  * @pdev: PCI device information struct
11180  *
11181  * Called to allow the driver to bring things back up after PCI error
11182  * and/or reset recovery has finished.
11183  **/
11184 static void i40e_pci_error_resume(struct pci_dev *pdev)
11185 {
11186         struct i40e_pf *pf = pci_get_drvdata(pdev);
11187
11188         dev_dbg(&pdev->dev, "%s\n", __func__);
11189         if (test_bit(__I40E_SUSPENDED, &pf->state))
11190                 return;
11191
11192         rtnl_lock();
11193         i40e_handle_reset_warning(pf);
11194         rtnl_unlock();
11195 }
11196
11197 /**
11198  * i40e_shutdown - PCI callback for shutting down
11199  * @pdev: PCI device information struct
11200  **/
11201 static void i40e_shutdown(struct pci_dev *pdev)
11202 {
11203         struct i40e_pf *pf = pci_get_drvdata(pdev);
11204         struct i40e_hw *hw = &pf->hw;
11205
11206         set_bit(__I40E_SUSPENDED, &pf->state);
11207         set_bit(__I40E_DOWN, &pf->state);
11208         rtnl_lock();
11209         i40e_prep_for_reset(pf);
11210         rtnl_unlock();
11211
11212         wr32(hw, I40E_PFPM_APM, (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
11213         wr32(hw, I40E_PFPM_WUFC, (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
11214
11215         del_timer_sync(&pf->service_timer);
11216         cancel_work_sync(&pf->service_task);
11217         i40e_fdir_teardown(pf);
11218
11219         rtnl_lock();
11220         i40e_prep_for_reset(pf);
11221         rtnl_unlock();
11222
11223         wr32(hw, I40E_PFPM_APM,
11224              (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
11225         wr32(hw, I40E_PFPM_WUFC,
11226              (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
11227
11228         i40e_clear_interrupt_scheme(pf);
11229
11230         if (system_state == SYSTEM_POWER_OFF) {
11231                 pci_wake_from_d3(pdev, pf->wol_en);
11232                 pci_set_power_state(pdev, PCI_D3hot);
11233         }
11234 }
11235
11236 #ifdef CONFIG_PM
11237 /**
11238  * i40e_suspend - PCI callback for moving to D3
11239  * @pdev: PCI device information struct
11240  **/
11241 static int i40e_suspend(struct pci_dev *pdev, pm_message_t state)
11242 {
11243         struct i40e_pf *pf = pci_get_drvdata(pdev);
11244         struct i40e_hw *hw = &pf->hw;
11245
11246         set_bit(__I40E_SUSPENDED, &pf->state);
11247         set_bit(__I40E_DOWN, &pf->state);
11248
11249         rtnl_lock();
11250         i40e_prep_for_reset(pf);
11251         rtnl_unlock();
11252
11253         wr32(hw, I40E_PFPM_APM, (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
11254         wr32(hw, I40E_PFPM_WUFC, (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
11255
11256         pci_wake_from_d3(pdev, pf->wol_en);
11257         pci_set_power_state(pdev, PCI_D3hot);
11258
11259         return 0;
11260 }
11261
11262 /**
11263  * i40e_resume - PCI callback for waking up from D3
11264  * @pdev: PCI device information struct
11265  **/
11266 static int i40e_resume(struct pci_dev *pdev)
11267 {
11268         struct i40e_pf *pf = pci_get_drvdata(pdev);
11269         u32 err;
11270
11271         pci_set_power_state(pdev, PCI_D0);
11272         pci_restore_state(pdev);
11273         /* pci_restore_state() clears dev->state_saves, so
11274          * call pci_save_state() again to restore it.
11275          */
11276         pci_save_state(pdev);
11277
11278         err = pci_enable_device_mem(pdev);
11279         if (err) {
11280                 dev_err(&pdev->dev, "Cannot enable PCI device from suspend\n");
11281                 return err;
11282         }
11283         pci_set_master(pdev);
11284
11285         /* no wakeup events while running */
11286         pci_wake_from_d3(pdev, false);
11287
11288         /* handling the reset will rebuild the device state */
11289         if (test_and_clear_bit(__I40E_SUSPENDED, &pf->state)) {
11290                 clear_bit(__I40E_DOWN, &pf->state);
11291                 rtnl_lock();
11292                 i40e_reset_and_rebuild(pf, false);
11293                 rtnl_unlock();
11294         }
11295
11296         return 0;
11297 }
11298
11299 #endif
11300 static const struct pci_error_handlers i40e_err_handler = {
11301         .error_detected = i40e_pci_error_detected,
11302         .slot_reset = i40e_pci_error_slot_reset,
11303         .resume = i40e_pci_error_resume,
11304 };
11305
11306 static struct pci_driver i40e_driver = {
11307         .name     = i40e_driver_name,
11308         .id_table = i40e_pci_tbl,
11309         .probe    = i40e_probe,
11310         .remove   = i40e_remove,
11311 #ifdef CONFIG_PM
11312         .suspend  = i40e_suspend,
11313         .resume   = i40e_resume,
11314 #endif
11315         .shutdown = i40e_shutdown,
11316         .err_handler = &i40e_err_handler,
11317         .sriov_configure = i40e_pci_sriov_configure,
11318 };
11319
11320 /**
11321  * i40e_init_module - Driver registration routine
11322  *
11323  * i40e_init_module is the first routine called when the driver is
11324  * loaded. All it does is register with the PCI subsystem.
11325  **/
11326 static int __init i40e_init_module(void)
11327 {
11328         pr_info("%s: %s - version %s\n", i40e_driver_name,
11329                 i40e_driver_string, i40e_driver_version_str);
11330         pr_info("%s: %s\n", i40e_driver_name, i40e_copyright);
11331
11332         i40e_dbg_init();
11333         return pci_register_driver(&i40e_driver);
11334 }
11335 module_init(i40e_init_module);
11336
11337 /**
11338  * i40e_exit_module - Driver exit cleanup routine
11339  *
11340  * i40e_exit_module is called just before the driver is removed
11341  * from memory.
11342  **/
11343 static void __exit i40e_exit_module(void)
11344 {
11345         pci_unregister_driver(&i40e_driver);
11346         i40e_dbg_exit();
11347 }
11348 module_exit(i40e_exit_module);